SUMMARY
This study seeks to advance research related to eXtensible Business Reporting Language (XBRL). XBRL is an open standard for reporting structured financial information which enables the efficient gathering of data and automated comparison of financial information. To encourage research using XBRL we describe the richness of XBRL data and sources from which it can be obtained. We follow with a review of the literature, beginning with research examining the adoption and use of XBRL by capital market participants. Next, we discuss data quality concerns that may impact the use of XBRL data, followed by a discussion of how auditors use XBRL data and their potential role in the assurance of the data. We then present literature that uses meta and underlying XBRL data to examine financial statement characteristics and disclosure properties. Based on the review of the literature, we identify topics with the greatest potential for future research.
I. INTRODUCTION
The objective of this study is to advance research related to eXtensible Business Reporting Language (XBRL), an open standard reporting language designed to improve the disclosure of financial information. This technology facilitates standardized reporting terminology, automated data extraction and usage, and the comparison of data across companies and over time.1 The introduction of XBRL is perhaps the most significant advancement in corporate financial reporting since the Securities and Exchange Commission (SEC) required companies to file financial reports electronically via the Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system in 1996. Emphasizing its potential, S. P. Kothari, Chief Economist and Director of the Division of Economic and Risk Analysis at the SEC, notes that “structured data will drive future research” (Kothari 2019).
In this review, we pursue our objective of advancing XBRL research in several ways. First, we introduce academic researchers to the unique, rich, and broadly available data, including how to obtain and use it. Second, we describe the background and history of XBRL and provide a review of the literature about XBRL adoption and different types of XBRL data. Our literature review focuses on empirical XBRL research relevant to the accounting and audit literatures, as identified in 99 working papers and articles published in 16 journals from 2004 through 2020.2 In addition, because the successful use of XBRL data depends on the accuracy of the data, we offer insights on the assurance of XBRL data and the potential role of auditors, standard setters, and researchers in maintaining and examining XBRL data quality.
Motivating this study are several factors that make XBRL data uniquely advantageous for academic research. XBRL filings contain the exact details reported in a company's financial statements and notes. This is unlike commercial databases (e.g., Compustat) that disseminate standardized, less granular data that may differ from amounts originally reported in the financial statements (Chychyla and Kogan 2015). Additionally, XBRL filings include all data in the notes to the financial statements which are not fully available from third-party data providers and are often hand-collected by users. For instance, while Compustat includes about nine fair value (FV)-related items, over 200 items are available from XBRL data (Ahn, R. Hoitash, and U. Hoitash 2020a). This example illustrates the granularity of the data available in XBRL filings and the potential for future research using this data. Finally, XBRL data are available for a broader set of firms than those typically available from commercial databases. To facilitate future research using XBRL data we provide a technical introduction to XBRL in Part II and Appendix C. Appendix C also identifies several XBRL data sources, each requiring varying levels of technical skills to utilize.
The literature review portion of this study begins with a presentation of studies examining the use of XBRL data by capital market participants. This research examined the types of firms and governance structures that chose to adopt XBRL and the impact of XBRL filings on the production of financial reports and financial markets. Many studies examined whether the primary objectives of XBRL, to reduce information processing costs and information asymmetry and thereby increase market efficiency, have been achieved. Other early research examined data quality concerns, documenting errors in XBRL filings such as inaccurate values (monetary, percent) and scale (thousands, millions). There is currently no requirement that XBRL data be audited, which we posit contributes to the likelihood of errors in XBRL data and to the risk of using XBRL.
The next section of our review discusses another measure of XBRL quality: whether financial statement preparers select the appropriate XBRL tags from the FASB taxonomy and/or appropriately create custom company-specific extended tags.3 We present a summary of studies examining firms' choice of extension use and discuss a subsequent body of literature that examines how the use of extended tags improves or mitigates capital market participants' abilities to compare information across companies and over time. These studies offer mixed results on the influence of extended tags on information processing costs and comparability.
Our next literature streams examine the use of tags to identify unique XBRL data. XBRL tags provide metadata, which is data that describes and provides information about other data.4 Often, rather than rely on the content reported in the tag (value or text), many researchers leverage the XBRL tag itself to explore their research questions. For example, recent research used XBRL tags to construct measures of financial reporting complexity (R. Hoitash and U. Hoitash 2018), comparability (Caylor, Chambers, and Mutlu 2019), and benchmarking (R. Hoitash, U. Hoitash, Kurt, and Verdi 2020). Other studies have used numerical (Aland and Burks 2020) or footnote textual data (Peterson, Schmardebeck, and Wilks 2015; Ahn, R. Hoitash, and U. Hoitash 2020b; Burke, R. Hoitash, U. Hoitash, and Xiao 2020) identified using XBRL tags. Using XBRL tags is advantageous because it provides access to large scale monetary and textual information that otherwise has to be manually collected.
We conclude our review with a discussion on how auditors may interact with XBRL data. First, auditors may use XBRL to streamline risk assessments and analytical procedures. Despite the potential for the benefits of XBRL data to extend to auditors, to date, limited research has examined the association between XBRL adoption and audit outcomes. We also consider auditors' role in the assurance of XBRL data. Current regulation does not require the assurance of XBRL filings, which has the potential to significantly undermine its usability due to potential quality concerns. The matter of potential assurance raises questions about appropriate sampling, materiality, and controls for audits of XBRL filings. Assurance challenges are further heightened with the introduction of iXBRL, in which unaudited XBRL data is embedded in audited HTML filings.5
We close our study highlighting opportunities for future XBRL research. While extant research has addressed various questions proposed in prior literature, developments in regulations and data quality create ample opportunity for further inquiry to inform auditors, regulators, financial statement users, preparers, and other capital market participants. To maximize the potential of XBRL's benefits for the capital markets, regulators should consider requiring the assurance of XBRL filings. We encourage researchers to continue investigating the costs and benefits of assurance of XBRL filings and propose potential research questions related to this issue. Additionally, future research may examine the impending Inline XBRL (iXBRL) mandate, which requires XBRL to be embedded in the financial statements filed in HTML format. However, we believe that the greatest opportunity for future research lies in the rich and unique data available from XBRL tags. The data is increasing in availability and reliability and offers information not available through traditional commercial databases. Studies may use tag metadata to construct measures that capture different properties of audited financial reports. Furthermore, the growing body of research on the textual properties of corporate filings may benefit from using XBRL tags that provide better accuracy in identifying disclosures than traditional methods of textual analysis.
The remainder of this paper is organized as follows. Section II describes the history of XBRL and provides a short technical overview of the data. Section III describes our literature review methodology. Section IV reviews the literature on XBRL adoption. Sections V, VI, and VII review the literature on XBRL tags, specifically extended tags, tag metadata, and the use of tags to identify numerical and textual financial statement data. In Section VIII we discuss auditors' use and assurance of XBRL. Section IX offers directions for future research, and Section X concludes.
II. XBRL: AN INTRODUCTION AND OVERVIEW
This section discusses the history of XBRL and the XBRL reporting structure and presents an introduction on how to obtain and use XBRL data.6
History of XBRL
In 2009, the SEC passed Interactive Data to Improve Financial Reporting, which required companies to report financial statement information in XBRL format. The timeline of XBRL adoption is presented in Figure 1.
XBRL adoption was initially voluntary. Starting in 2005, the XBRL Voluntary Filing Program (VFP) allowed firms the choice of submitting financial statements in XBRL format (SEC 2005). The VFP was intended to provide participants with the opportunity to practice the XBRL reporting process and to provide the SEC with information that could better inform the mandate that followed the voluntary period. Mandated XBRL reporting for Form 10-K was phased in based on accelerated filer status.7 In the initial year of compliance, filers were required to tag individual line items in the face of the financial statements (e.g., income statement). There was no requirement to tag numbers in the notes to the financial statements and attached schedules, rather, each was tagged as a single block of text.8 Beginning the second year of compliance, in addition to tagging each quantitative disclosure in the face of the financial statements, filers were required to tag each quantitative disclosure within the notes to the financial statements or schedules.
In 2018 the SEC adopted an amendment that requires firms to file Inline XBRL (iXBRL) reports in which XBRL is directly embedded into HTML filings. Thus, instead of submitting two filings (HTML and XBRL), financial statement preparers now submit a single set of documents. This change has several objectives, including a reduction of the burden on preparers by eliminating duplication of filings, giving preparers full control over the presentation of XBRL disclosures, enhancing the usability of these disclosures, and improving the quality of filings by making it easier to identify errors (Basoglu and White 2015). The iXBRL requirement will be phased in over a three-year period based on accelerated filer status beginning with fiscal years ending on or after June 15, 2019, but some firms voluntarily filed in iXBRL format before the mandate.9
Technical Overview of XBRL
XBRL is an XML-based markup language used to electronically communicate financial and business data. A markup language uses tags to encode data. When filing XBRL reports, financial-statement preparers tag each financial concept in Item 8 of the 10-K (and interim 10-Q) filings.10 This tagging applies to the financial statements, such as the balance sheet and income statement, as well as to the notes to the financial statements. The tags in turn enable the automatic extraction, sorting, and comparison of information across firms and over time.11 An illustration of XBRL tags and iXBRL tags can be found in Appendices A and B, respectively.
In addition to numerical values, XBRL tags are designed to capture textual properties of the financial statements and notes. Such tags are often referred to as TextBlocks, as they capture blocks of text. A TextBlock XBRL tag can be used to capture an entire footnote, an accounting policy, or a table in the notes to the financial statements, each of which can be used to perform textual analysis.12 We provide further details on how to use XBRL data in Appendix C. Appendix C also provides information about XBRL data sources, including the relative difficulty of using each type of data source: custom computer programs, SEC datasets, researcher processed data, XBRL US, and other third-party XBRL data providers.13
Both numerical and TextBlock tags may be categorized as either taxonomy or extended tags. Taxonomy tags are pre-defined tags representing accounting concepts that appear in the XBRL U.S. GAAP Taxonomy. The Taxonomy is a machine-readable “dictionary” that is updated annually and may be downloaded from the FASB website at http://xbrl.fasb.org.14 Taxonomy tags are indicated by the prefix “us-gaap.” For example, <us-gaap:Revenues> is a Taxonomy tag which represents the amount of recognized revenue. Each tag can appear multiple times in a financial document. For example, the tag <us-gaap:Assets> typically appears several times with different values, representing values for different business units and time periods. Each tag contains metadata, such as whether the tag is monetary, a percentage, textual or other, the period that the value refers to, whether the value is debit or credit, and more.15 XBRL metadata allows users of the data to distinguish between each unit and time period in which a tag is used. The value of each tag appearance is termed a fact.16 The facts and tags allow for the extraction of more detailed financial information, some of which is not available in traditional commercial databases.
To fully support the objective of XBRL to increase automatic cross-sectional benchmarking, companies are required to use taxonomy tags if one is available. However, the taxonomy may not always meet the precise reporting needs of all firms, so companies may extend the taxonomy and create their own extended tags. These tags are identified by company ticker symbols as a prefix instead of the “us-gaap” prefix. Importantly, while extended tags allow companies to make decisions about how to best represent their performance, extended tags potentially reduce the efficacy of cross-sectional analysis.17 Appendix A presents examples of both taxonomy and extended XBRL tags.
III. LITERATURE REVIEW METHODOLOGY
Our journal selection was motivated by identifying those with the greatest impact in accounting literature as well as those that frequently publish XBRL research (Bedard, Deis, Curtis, and Jenkins 2008; Carcello, Hermanson, and Ye 2011; Lennox and Wu 2018).18 In addition, because XBRL research is experiencing recent growth in academic inquiry, we deemed it appropriate to include working papers in our review.19Table 1 highlights XBRL working papers and research published in top-tier accounting journals and the Journal of Information Systems.20 Within the identified sources, we conducted our search for articles containing the terms “XBRL,” “eXtensible Business Reporting Language,” “Interactive Financial Information,” “XBRL for Interactive Data” and “Interactive Data.”21,22
Table 1 reports the number of XBRL-related papers by journal and year. The Table illustrates that papers published in the top accounting journals are concentrated in the last three years, suggesting that XBRL research is gaining traction in its ability to provide opportunities for future publications in accounting journals. More importantly, the increase in publications over time suggests that academia is recognizing the value and contributions of research using XBRL data. We categorize the literature into five themes (adoption; extended tags; metadata; numerical and footnote data; audit & assurance), presented by year and journal in Table 2, Panels A and B, respectively. Table 2, Panel C presents the literature classified into each category.23 Interestingly, studies using metadata are concentrated in the last five years of the sample period, suggesting an opportunity to further leverage this data. Fewer studies have used numerical and footnote (textual) XBRL data, perhaps because using the data is more challenging than using raw XBRL tags.24
IV. XBRL ADOPTION
This section discusses literature examining the influence of XBRL on the capital markets after voluntary and mandatory adoption. We also present XBRL data quality concerns that arose during this timeframe.
Characteristics of Participating Firms
The voluntary filing period gave initial insight into firms' characteristics, willingness, and ability to adopt XBRL. Firms participating in the VFP have a higher propensity to voluntarily disclose information, higher profitability (Boritz and Timoshenko 2015), are larger (Premuroso and Bhattacharya 2008; Callaghan and Nehmer 2009), and less financially leveraged (Callaghan and Nehmer 2009). For non-high-tech firms, innovativeness is also an explanatory factor for VFP participation (Boritz and Timoshenko 2015). The literature reports mixed results between analyst following, Big 4 auditors, and high earnings quality and the likelihood of participation in the VFP (Boritz and Timoshenko 2015). Firms in less competitive industries are more likely to voluntarily adopt XBRL due to the increase in proprietary costs related to firms' voluntary disclosures (Chatterjee, Gupta, and Kong 2020).25
Some evidence suggests that firms with strong corporate governance are more likely to voluntarily adopt XBRL reporting due to stronger institutional mechanisms that drive improved firm disclosures (Premuroso and Bhattacharya 2008; Boritz and Timoshenko 2015). In contrast, other research finds that firms with weaker corporate governance are more likely to participate in the VFP to signal to the market that their institutional mechanisms are strong (Callaghan and Nehmer 2009). Many voluntary adoption studies compare voluntary XBRL adopters with non-adopters. However, this literature does not consider differences in the quality of XBRL reporting between adopting firms and therefore it is difficult to disentangle between firms that adopted to satisfy stakeholders and those with true commitment to this new reporting technology.
Does XBRL Reporting Benefit Capital Market Participants?
The SEC contended that “interactive data has the potential to increase the speed, accuracy, and usability of financial disclosures” (SEC 2009) and thus lower information processing costs. Research offers mixed evidence on whether the availability of XBRL information resulted in a more timely and efficient market response. Du and Wu (2018) find evidence of a shorter reporting lag for annual and quarterly filings for both large accelerated and accelerated filers, but not for non-accelerated filers.26Efendi, Park, and Smith (2014) document improved information efficiency post-XBRL mandate via a reduced post earnings announcement drift. Examining the usability of information, Dhole, Lobo, Mishra, and Pal (2015) find that financial statement comparability declined after the XBRL mandate.
The XBRL mandate also proposed that providing financial information in digital format should allow investors to collect and analyze data in a more efficient manner by leveraging automated data collection techniques. XBRL adoption studies draw conclusions about the intended reduction in information processing costs based on observed market outcomes. There is generally a lack of evidence to suggest that filings made during the VFP reduce information processing costs, perhaps due to the lack of reliability of XBRL reports filed during this period (Dong, Li, Lin, and Ni 2016). Some mandatory adoption research suggests detrimental effects of XBRL on information processing costs, as evidenced by reduced analyst forecast accuracy (Liu, Wang, and Yao 2014a; Felo, Kim, and Lim 2018).27 However, many studies suggest that XBRL mitigates investors' information processing costs and thus reduces information asymmetry (Blankespoor, deHaan, and Marinovic 2020). This is evidenced by lower stock return synchronicity (Dong et al. 2016), lower return volatility (Kim, Lim, and No 2012; Huang, Shan, and Yang 2020), reductions in bank loan spreads (Chen, Kim, Lim, and Zhou 2018b), lower investor expectations of crash risk (Zhang, Guan, and Kim 2019), and a positive market reaction to news related to XBRL legislation (Chen, Wang, and Zhou 2018a) and XBRL filings (Hao and Kohlbeck 2013).28,29
Another major goal of XBRL was to allow automated financial analysis. The SEC (2009, 1) states, “In this format, financial statement information could be downloaded directly into spreadsheets, analyzed in a variety of ways using commercial off-the-shelf software, and used within investment models.” Accordingly, studies have examined whether improved information processing extends to machines. The evidence is conflicting, with Pungaliya and Wang (2020) finding a positive association between machine processing and stock returns post-XBRL mandate, and Allee, DeAngelis, and Moon (2018) failing to find evidence that XBRL improves computerized information processing.
One of the inherent limitations of studies examining market outcomes of XBRL adoption is the assumption that investors, and other capital market participants, are actually using XBRL data. This issue was of strong concern during the voluntary adoption period, during which XBRL data was a novel source of information, and firm participation in the VFP was limited. Indeed, financial officers expressed doubts that investors were using XBRL tags (Janvrin and Mascha 2014). However, evidence suggests the XBRL mandate improved information accessibility (Yen and Wang 2015) and acquisition (Chen and Zhou 2019), and thus the use of XBRL data should have increased since the VFP.
The ability to access detailed XBRL data on the EDGAR system, rather than from third-party providers such as Compustat, should make data available to a broader set of investors. While concerns about the use of XBRL data declined during the mandatory adoption period, the question remained of whether the benefits of XBRL extend to all market participants or whether use of the data is primarily by large investors and institutions (Harris and Morsfield 2012).30 Studies offer mixed evidence on whether XBRL was successful in mitigating the information gap between large and small investors. Some literature offers support (Kim, Li, and Liu 2019b) while other research finds an increase in the information gap following the adoption, likely attributable to large investors' superior resources and abilities to process information filed in XBRL (Blankespoor, Miller, and White 2014; Cong, Hao, and Zou 2014).31,32 Offering further juxtaposing conclusions, users of small company filings are able to access XBRL files and prefer them to non-XBRL files (Cong, Du, and Vasarhelyi 2018).33
In sum, extant literature documents mixed results on the impact of XBRL filings on the capital markets. The conflicting conclusions suggest that the effects are complex, differing across participants, time, and market characteristics.
Quality Concerns: How Well are Firms Implementing XBRL?
The body of research examining how capital market participants respond to XBRL data assumes that the data was accurate and appropriate. In this section, we discuss data quality concerns arising from the implementation of new reporting practices.
Financial statements are prepared and certified by management and audited by the external auditor. Historically, the financial statements were converted to HTML format and this file was provided to the SEC. After the XBRL mandate, firms were required to furnish filings in both HTML and XBRL formats. Although the interactive XBRL filings (financial statements and notes) should theoretically be identical in content to the original financial statements and traditional HTML filings, there is no mandate that these filings receive any form of assurance from the external auditor. As a result, the SEC distributes two sets of filings, one audited (HTML) and one unaudited (XBRL). This approach is concerning because evidence suggests that many XBRL filings are inconsistent with the HTML files or contain errors (Boritz and No 2008, Debreceny, Farewell, Piechocki, Felden, and Gräning 2010; Farewell, Hao, Kashyap, and Pinsker 2017).
The conversion from HTML to XBRL creates pitfalls for errors in metadata and file structures. One such threat to XBRL data is errors in metaproperties assigned to each tag. For example, Calcbench (2014) finds that the XBRL filings of a significant number of firms contain at least one scaling error. Values presented in thousands instead of millions can lead to errors in models, decision making, and trading. Finally, incorrect file structures may negatively impact the usability of XBRL files, such as having the incorrect nesting or presentation that prevents users from linking the value of XBRL tags to their financial statements and notes.
Several papers suggest that the frequency of firms with XBRL errors has significantly declined (Bartley, Chen, and Taylor 2011; Debreceny et al. 2011; Du, Vasarhelyi, and Zheng 2013). This decline may be attributable to preparers' learning curves as well as improvements in XBRL oversight. In 2018, the SEC introduced a version of EDGAR that generates automatic messages to flag errors in XBRL filings (SEC 2018). XBRL submissions are now flagged when items are incorrectly tagged as negative, when outdated tags are used, and in some cases if a company submits an extended tag when a standard taxonomy tag already exists. Of course, XBRL errors that cannot be flagged will persist. These errors may include tags related to complex accounting standards or tagging of financial statement footnotes. The assurance of XBRL filings has the potential to reduce the risk of errors and increase the credibility and usability of XBRL filings.
V. EXTENDED XBRL TAGS
Extended XBRL tags allow companies the flexibility to accurately depict their financial information. Some companies, however, may use extended tags inappropriately, either due to misunderstanding of the XBRL taxonomy, or to intentionally obfuscate financial reports. Such errors generate concerns about the reliability of XBRL data arising from managers' decisions to use extended tags (Whitehouse 2011; Harris and Morsfield 2012). Indeed, a significant number of XBRL extensions are created when an equivalent tag is already provided in the taxonomy (Boritz and No 2009; Debreceny et al. 2011). Scherr and Ditter (2017) separate necessary from unnecessary extensions and examine what firm characteristics drive XBRL extension use. They find that the use of necessary extended tags is associated with the complexity of financial reporting and costs of voluntary disclosure.34 Unnecessary extension use is most common in firms with less experience in XBRL reporting or less involvement in the XBRL tagging process. Overall, their results suggest that the use of taxonomy extensions is related to firms' attempts at accurate financial reporting, rather than management's discretion to hinder information processing. The SEC also examines extended tag use (Cohn 2016; SEC 2019), potentially deterring managers from using them for obfuscation.35
Figure 2 illustrates the average percentage of extended tag use from 2012 to 2019.36 The data indicates a decline in extension use over the sample period, with high use of extensions in the first year of mandatory adoption, approximately 16 percent, declining to an average of approximately 14 percent in later years.
The decline in extension use may be attributable to a learning curve, in which financial statement preparers are better able to identify the appropriate taxonomy tags, and to the introduction of more comprehensive taxonomy tags by the FASB.
Regardless of whether XBRL tags are appropriate or duplicative of taxonomy tags, extensions may reduce the usability, extractability, and comparability of XBRL filings by limiting users' abilities to identify and compare specific line items or footnote disclosures. The literature offers mixed results on the impact of the potential reduction of usability and comparability on investors' information processing. Some research suggests that extended tags improve analysts' information processing, as evidenced by reduced forecast dispersion, increased forecast accuracy (Johnston 2020; Li and Nwaeze 2018), and increased analyst following (Kirk, Vincent, and Williams 2016; Li and Nwaeze 2018).37 In contrast, other studies suggest that extensions mitigate analysts' abilities to compare information across companies and over time, as evidenced by a positive association between extensions and analyst forecast dispersion and a negative association with analyst forecast accuracy (Kirk et al. 2016; Felo et al. 2018; Henry, Liu, Yang, and Zhu 2020).
If XBRL is successful in mitigating users' information processing costs, capital market participants may be better able to process financial statement information (Hodge, Kennedy, and Maines 2004) and identify earnings management activities. Standardized tagging of financial statement elements may aid investors in analyzing financial reports. The intention of extended tags is to provide management with discretion in presenting XBRL information in the manner that most accurately reflects the originally prepared financial reports. However, this discretion may mitigate investors' abilities to monitor earnings management behavior. The evidence on this issue from the extant XBRL literature generally supports the latter hypothesis, documenting a negative association between the use of extended tags and financial reporting quality (Scherr and Ditter 2017; Hoitash and Hoitash 2018; J. Kim, J. W. Kim, and Lim 2019a).38
Other literature considering the effects of customized extension use on information processing costs finds mixed results on the benefits of extended tags. Allee et al. (2018) fail to find evidence that XBRL quality, based on various considerations of extension use, measures the machine readability of financial reports. The use of extended tags increases loan spreads for banks post-XBRL adoption (Chen, Kim, Lim, and Zhou 2018b), increases stock return synchronicity (Lim, Richardson, and Smith 2020), and increases investors' expectations of crash risk (Zhang et al. 2019). Alternatively, Kim, Li, and Liu (2019b) find evidence to support the notion that XBRL, and the use of extended tags, reduces information processing costs, making for a more transparent information environment, and thus firms reporting in XBRL attract more shareholders. In sum, the impact of extended tags on financial reporting quality and subsequent interpretation of financial information by capital market participants is inconclusive, motivating the need for further research on the subject.
VI. METADATA IDENTIFIED FROM XBRL TAGS
This section reviews the literature using XBRL metadata, data that provides information about another source of data. XBRL tags provide metadata on the actual monetary or textual values contained in financial reports.
Using Metadata to Identify Financial Statement Characteristics
Hoitash and Hoitash (2018) count the number of XBRL tags in 10-K filings to develop a measure of accounting reporting complexity (ARC). They propose that companies with more diverse economic activities require more accounting concepts to describe these activities in the financial statements and thus, more XBRL tags. Relative to other commonly used measures of complexity, they document that this measure of ARC is more consistently associated with a greater likelihood of misstatements and material weaknesses, longer audit report lag, and higher audit fees.39 The associations using ARC also have greater explanatory power and economic significance than other proxies for complexity. Likewise, they document an inverse association between the number and percent of extended tags with these outcomes.40
Subsequent literature extends the ARC measure to the account-level. Brown, Cohen, and Huffman (2019) leverage the ability of ARC to capture account-specific complexity and find that the number of non-GAAP disclosures increases with the reporting complexity of specific accounts such as derivatives and income taxes. R. Hoitash, U. Hoitash, and Yezegel (2019) use XBRL to develop measures of account-specific complexity and financial analysts' account-specific expertise. They use detailed XBRL disclosures to construct analysts' account-specific expertise in fair value, derivatives, and pensions and find that this form of expertise can mitigate the negative effects of complexity associated with these accounts.
Chychyla, Leone, and Minutti-Meza (2019) follow the construction of ARC and further link each XBRL tag to the relevant standard for that line item to measure the length of the accounting standard associated with each tag. Thus, their measure of financial reporting complexity (FRC) captures the complexity of accounting standards in combination with the number of disclosed accounting items. They document a positive association between FRC and the level of accounting expertise on the board and audit committee, and this expertise can mitigate the relationship between FRC and negative reporting outcomes such as restatements and disclosures of material weaknesses.
Several studies have used XBRL to develop measures of financial reporting benchmarking and comparability using a broad set of accounting information. Hoitash et al. (2020) construct an XBRL-based measure of pairwise financial statement benchmarking (FSB). The measure identifies pairs of firms for benchmarking and also identifies biased benchmarking choices by practitioners, such as corporate boards and analysts. Furthermore, they show that firms with greater benchmarking scores are associated with improved analyst forecast accuracy and reduced forecast dispersion. Considering financial reporting similarity in the audit context, Johnston and Zhang (2020) count the number of items in XBRL filings to develop a measure of financial reporting similarity and find that firms that share the same auditor have more similar accounts reported on the face of the financial statements. Henry et al. (2020) use the Cosine similarity construct as a measure of structural comparability. They find that earnings comparability increases analyst coverage and analyst forecast accuracy and decreases analyst forecast dispersion through structural comparability. Caylor et al. (2019) use the same measure of the mean of a firm's pairwise overlap of data items with peer firms, which they refer to as financial statement uniformity. The authors note using XBRL data allows them to create a measure of uniformity for the entire set of financial statements, including the notes, which Compustat does not enable.
Using Metadata to Identify Firm Disclosure Changes
XBRL metadata is also used to identify changes to firm disclosures in response to regulators. Bozanic, Hoopes, Thornock, and Williams (2017) examine how public and private disclosures interact to influence tax regulator enforcement and firm disclosure. They use XBRL data in supplemental analysis to document changes in the number of tax-related tagged numbers and find that when the proprietary costs of disclosure to the IRS are lower, firms disclose more tax-related quantitative information. Ahn et al. (2020a) use XBRL data to measure the changes in FV-related tags and find that firms increase the number of FV-related tags following receipt of a comment letter from the SEC. Using XBRL tag metadata, Ahn et al. (2020a) were also able to distinguish Level 3 FV tags from tags capturing FV Levels 1 and 2.
VII. NUMERICAL AND TEXTUAL DATA TAGGED WITH XBRL
This section describes studies using XBRL to obtain financial statement numerical and footnote (textual) information that previously had to be manually collected.
Numerical Data
One of the benefits of XBRL is that the data represents information reported directly in the financial statements, rather than data that has been standardized or aggregated by third-party providers (e.g., Compustat). An example of leveraging the benefits of numerical XBRL data are provided by Henselmann, Ditter, and Scherr (2015), who use XBRL data to extract monetary line items and identify firms suspected of engaging in earnings management based on abnormal digit distributions. Another example is provided by Aland and Burks (2020), who use XBRL to identify the gross gain and gross loss components of banks' net realized gain or loss, the breakdown of which is not available from data aggregators.
Textual Data
XBRL tags may also be used to study specific footnotes or general characteristics of textual disclosures. Table 3 presents a list of the 12 most frequently used tags capturing complete financial statement footnotes, as well as the percentage of firms that use these tags in their 10-K filings.
Several studies extract detailed footnote disclosures using XBRL TextBlock tags and perform textual analysis to make inferences about audit outcomes. Czerney and Sivadasan (2020) use XBRL tags to identify footnote disclosures in the financial statements and find that external auditors, rather than management, have the greatest influence on firms' reporting of textual footnote disclosures. Czerney, Lisic, Wu, and Zhang (2019) use XBRL-identified footnotes and find that the tone of disclosures is more reflective of bad news for companies using a Big 4 auditor than a non-Big 4 auditor. Using FV-related line items, Ahn et al. (2020a) document that firms increase the number of words in the FV footnote following receipt of a comment letter from the SEC and that the increase in the number of words is greater among clients with FV expert auditors. Burke et al. (2020) use XBRL to identify and examine footnotes referenced in critical audit matters (CAMs) and find that footnote disclosures get longer, less sticky, and more uncertain following a CAM reference.
Other studies exemplify the variety of footnotes that may be obtained using XBRL. Ahn et al. (2020b) use XBRL to identify the fair value, goodwill and intangibles, and business combinations footnotes and construct disclosure quality measures of textual relative to numerical intensity in each footnote. Peterson et al. (2015) use XBRL to identify the accounting policy footnote and create a text-based similarity measure to capture firms' accounting consistency. They document a positive association between accounting consistency and earnings quality and conjecture that using XBRL increases the accuracy in capturing this particular footnote. Inger, Meckfessel, Zhou, and Fan (2018) use XBRL to extract tax footnotes and calculate the readability of tax disclosures. They document a positive (negative) association between tax avoidance and tax footnote readability for firms with tax avoidance below (above) the median. Bozanic et al. (2017) use XBRL to identify tax footnotes and examine changes in disclosures as well as the number of uncertain items identified in the tax footnote. Also using XBRL to obtain tax information, Schwab, Stomberg, and Xia (2020) use XBRL to identify tax rate reconciliation tables and examine factors contributing to tax avoidance.
VIII. AUDITORS' USE AND ASSURANCE OF XBRL
Here we discuss how auditors may interact with XBRL data. We address the potential for XBRL to improve the audit process and then present challenges of assurance of XBRL and iXBRL data.
How Are Auditors Using XBRL?
While the intended beneficiaries of mandatory XBRL reporting are investors, other stakeholders, including auditors, may also benefit from the use of XBRL data.41 In order to opine on the financial statements, auditors must complete a series of audit procedures. The audit process includes risk assessments and analytical procedures, both of which require significant data analysis. The structured nature of XBRL data allows for streamlined acquisition of accounting data (Debreceny and Gray 2010; Gray and Debreceny 2014). Thus, auditors are better able to benchmark and compare audit clients with peer firms and prior-year reporting for purposes of identifying anomalies and risks (Gambetta, García-Benau, and Zorio-Grima 2016).
XBRL data includes numbers presented in the face of the financial statements (e.g., income statement), as seen in Appendix A, or supplementary information contained in the notes to the financial statements that is otherwise hard to collect. For example, auditors may use XBRL to collect peer footnote data to analyze the effect of the expected rate of return on pension valuations and net income. An additional benefit of XBRL data to auditors is that it is available in real time, as published by public companies, providing immediate access to peer companies' information. Finally, as suggested by the SEC's Office of Structured Disclosure, auditors may be able to leverage the structured nature of XBRL data to better enable visualizations of narratives and disclosures (Willis 2019).
While XBRL has significant potential to benefit auditors, there has been limited research on the association between XBRL adoption and audit outcomes. Amin, Eshleman, and Feng (2018) suggest that XBRL facilitates streamlined data acquisition and analysis, improves internal controls, and creates opportunities for continuous auditing. They therefore test for efficiencies in the audit process and find that the XBRL mandate is associated with a decline in audit report lags. The results are concentrated among clients in high-tech industries and those with strong internal control systems, but are lessened for clients of Big 4 audit firms. They find further evidence of a learning curve in XBRL reporting and a negative association with audit fees.
Assurance of XBRL
Current regulations do not require an audit of XBRL filings, which may undermine the usability of the data.42 The audit of detailed XBRL data, including the numbers presented in XBRL, XBRL tags, metadata, and file structures, may thus offer benefits to the capital markets. Potential assurance of XBRL data, however, is not without challenges. Auditors opine on the financial statements as a whole, indicating whether the financial statements accurately present a company's financial position. Auditors do not opine on individual line items or disclosures within the financial statements. Alternatively, since the intended usage of XBRL reports is of individual data identified from XBRL tags, rather than the entirety of the financial statements, the assurance of these reports should potentially also be at the data-level (Gunn 2007). In other words, assurance of XBRL filings may best serve users if auditors opine on the appropriateness of tag choices rather than the complete XBRL file or preparation process.
Due to the granularity of assurance needs, questions arise about appropriate levels of sampling in audits of XBRL tags (Plumlee and Plumlee 2008; Srivastava and Kogan 2010). Traditional sampling objectives involve determining materiality. In the context of XBRL, materiality could be based on material misstatements resulting from the tagging process (the dollar value represented in the tag) or on the number of tagging errors. Ignoring the dollar value may result in auditors exerting effort on a large volume of XBRL tags with low dollar value materiality thresholds. Additional audit considerations include control risks, such as whether traditional internal controls over financial reporting are sufficient to ensure the integrity of XBRL filings (R. Plumlee and M. Plumlee 2008; Srivastava and Kogan 2010).43 Finally, the lack of regulatory penalties for poor quality XBRL filings may contribute to low demand for assurance by filers. While the SEC reviews XBRL filings, we are not aware of any sanctions issued by the SEC for poor quality XBRL filings.
Absent a regulatory requirement for assurance over XBRL filings, research has considered other factors influencing the supply of and demand for such assurance.44 Survey evidence suggests that auditors are not interested in providing assurance over the XBRL process, likely due to liability concerns (Janvrin and No 2012). Financial statement preparers' perceptions of the benefits of assurance do not always exceed the perceived costs (Daigle and Lampe 2004). Therefore, reducing the perceived cost of external assurance may be an important consideration for preparers of XBRL filings (Alles and Gray 2012).
Additional Challenges Presented by Inline XBRL
The SEC's adoption of advancements to XBRL reporting brings further complications to assurance considerations. iXBRL requires firms to file a single report in which XBRL is directly embedded into the HTML filing. Financial statement users are able to click on tagged items to see the XBRL data associated with each distinct fact. In essence, the number they see is audited, but the underlying number and the metadata information revealed when users click on this amount is not.45 The embedded filings have thus generated concerns about perceived assurance over XBRL disclosures. EY (2017, 2) states: “We are concerned that embedding tags in the financial statements and having the SEC provide an iXBRL viewer to highlight them and display aspects of the tag could lead investors to assume that the tagging has been audited or reviewed by the registrant's independent, registered public accounting firm. Investors and other financial statement users also might assume that iXBRL tags were subject to the registrant's internal control over financial reporting (ICFR) on which the auditor issued an attestation report.” The lack of assurance of XBRL filings may expose auditors to additional litigation risk and cause investors to over-rely on unaudited XBRL data.
In summary, while the capital markets generally recognize the importance of reliable XBRL reporting, audited XBRL information is not required. The arguments we present above should encourage regulators and auditors to consider mandating and implementing XBRL audits.
IX. DIRECTIONS FOR FUTURE RESEARCH
XBRL offers a unique opportunity for academia to guide practice and regulation (Debreceny et al. 2005). Accordingly, this section identifies XBRL-related research opportunities. As the disclosure requirements mature and XBRL data becomes available from more sources, we encourage researchers to leverage XBRL tags and use innovative research designs to answer questions that cannot be answered with other data sources. While we have identified some areas for academic inquiry, this is by no means an exhaustive list of the growing potential of XBRL-based research.
Investor Use of XBRL
The use of XBRL data by capital market participants has increased over time. However, financial statement preparers may not have a sufficient understanding of how their financial statements are used by analysts and investors (Matherne 2019). Research may continue to consider how investors use and interpret XBRL-tagged data. For example, requests for EDGAR filings increased after XBRL adoption (Chen and Zhou 2019). Researchers may further examine how investors access XBRL data. We are not aware of evidence that identifies to what extent each of the common XBRL data sources are used by different types of investors. Future research may also examine differences in data quality across XBRL data providers and consequences of these differences for investors' use in decision-making. Furthermore, value may be added by research that documents whether mistakes and inconsistencies in XBRL filings are detrimental to investors, particularly to machine processing and automatic algorithmic trading. Behavioral research may consider how investors respond to the use of extended tags and what factors influence their interpretation of this information (Geerts, Graham, Mauldin, McCarthy, and Richardson 2013).
XBRL Meta-, Numerical, and Textual Data
Studies using XBRL tags to identify meta-, numerical, or footnote data are becoming increasingly prevalent and account for the majority of recent XBRL-related publications in top accounting journals. We believe that access to this data, which previously had to be hand-collected, will continue to fuel future XBRL research in audit and accounting. To assist academic researchers in determining the best methods by which to obtain XBRL data, Table 4 identifies XBRL data sources in extant literature. Panels A and B present the sources of data identified in papers using XBRL metadata and numerical and textual data, respectively. The majority of these papers (22) obtain the data via direct download by researchers.
Future studies may leverage XBRL tags to identify more account-specific contexts. For example, research can examine accounts such as debt, leases, or commitments and contingencies, and address questions that relate to specific contexts and outcomes. XBRL tags may facilitate a cost-benefit analysis of firms' choices to prepare XBRL filings, be it via in-house capabilities, software packages from service providers, or fully outsourced processes, and the associated quality considerations of each.
The use of numeric or textual data obtained from XBRL tags, particularly data presented in the notes to the financial statements, also presents significant opportunities. This data is often not included in conventional databases and using XBRL can facilitate efficient data gathering. Furthermore, XBRL data reflects information exactly as reported in the financial statements, which presents an opportunity to examine information that is lost when data is aggregated and standardized by commonly used third-party data providers. Additionally, executing textual analysis to perform natural language processing (NLP) on specific financial statement notes presents an excellent opportunity for future research because XBRL enables the accurate extraction and cross-sectional comparison of text in the footnotes. Thus, via both numeric and textual disclosures, XBRL increases the confidence that cross-sectional comparisons use comparable items.
Auditing and Assurance
The mandatory adoption literature generally examines if and how investors use XBRL. It remains an empirical question whether and how auditors use XBRL. The structured nature of XBRL data creates an opportunity for streamlined risk assessments and analytical procedures, but whether these specific processes improve post-XBRL adoption has not been tested. Research has used XBRL metadata to construct measures of financial statement characteristics like complexity, a measure of company risk. Auditors may be able to leverage similar measures to assist with audit pricing or planning efforts. Research may further consider whether extended tags impair auditors' abilities to use XBRL for analytical procedures or peer benchmarking.
Research yields inconclusive results about the impact of extended tag use on information processing. This lack of conclusion is unsurprising because extended tags currently capture multiple concepts. If extended tags are used correctly, they can capture a company's complexity and unique accounting needs. If used incorrectly, they capture either errors or obfuscation. While some research has attempted to bifurcate the two (e.g., Scherr and Ditter 2017), this subject offers additional evidence supporting the need for assurance of XBRL filings. If extended tags are audited for appropriateness, research may be able to identify more reliable evidence on the impact of extended tag use.
We thus recommend that researchers continue to examine the costs and benefits of assurance, such as audits of specific tag properties, tag use, or the overall preparation process (Gunn 2007). Surveys and experiments are particularly suitable to address this question. For example, surveying auditors, other assurance parties, and clients may shed light on the potential costs of assurance. Researchers may also consider whether auditors are the appropriate entities to provide assurance for XBRL-based financial reports (Vasarhelyi, Chan, and Krahel 2012), or whether internal audit, software providers, or other third-parties may have more appropriate technical expertise for the task. Under the current reporting standard, behavioral researchers may examine whether jurors hold auditors accountable for errors in XBRL filings absent an assurance requirement.
Other assurance-based research may identify firms that voluntarily choose to assure their XBRL reports and assess the incremental costs of such assurance.46 Within such a sample, research may also examine quality improvements such as the potential reduction in numerical errors, sign switches, scale errors, and the misuse of extended tags. Finally, future research may consider whether stronger regulatory penalties for poor XBRL quality result in higher quality XBRL reports. Combined, results from such studies may be informative to regulators considering the costs and benefits of XBRL assurance.
Inline XBRL
Regulations about XBRL disclosures continue to evolve, with the most recent change being the shift to iXBRL. Future research may examine the voluntary and early adopters of this technology to help inform the SEC on the success of iXBRL. This research may examine the quality of filings, characteristics of participating firms, and the impact of iXBRL reports on the capital markets.47 Future research may further examine whether iXBRL data allow investors to make better decisions (Vasarhelyi et al. 2012). To inform regulators on the consequences of iXBRL reporting, research may examine whether investors perceive the information presented in iXBRL reports to be assured by the external auditor and the consequences of this perception on users' decision making. Other audit-related topics for future research include the impact of iXBRL on audit outcomes. Does this reporting requirement influence the time it takes to complete an audit, as reflected in audit report lags or audit fees? Finally, while studies during the voluntary adoption period were limited due to uncertainty about the level of use of XBRL data, this concern is lessened for iXBRL adoption because XBRL is widely available and use of the data is prevalent. However, researchers may look for opportunities to identify unique features of iXBRL that offer direct evidence of use of this new reporting format.
X. CONCLUSION
This study reviews extant XBRL research, summarizes its contributions to the accounting and audit literature, and offers suggestions for future research. The goal of this synthesis is to inform researchers on the potential for XBRL data to answer previously unattainable research questions. We also seek to provide insights to stakeholders such as regulators and financial statement users and preparers about the costs and benefits of XBRL and XBRL assurance. We do so via a comprehensive review of XBRL literature, both peer-reviewed and working papers, focusing on XBRL adoption and the use of XBRL tags. The literature identifies benefits of XBRL reporting to market participants as well as potential threats to these benefits due to the lack of assurance requirements for XBRL reports.
To achieve our primary goal of introducing researchers to this rich data source, we have highlighted streams of literature that leverage XBRL tags in varying ways. As XBRL data is inherently technical in nature, to further our goal and assist accounting researchers in future inquiry using this data, we provide information on XBRL data sources, examples of XBRL and iXBRL filings from the SEC, and instruction for using XBRL data based on practical experience. We also point researchers to shared XBRL data on accounting reporting complexity (ARC) as well as textual footnote data in a designated website: xbrlresearch.com. We encourage other researchers using XBRL to demonstrate a collaborative spirit and share databases and other XBRL resources with the academic community.
REFERENCES
APPENDIX A
An Illustration of Taxonomy and Extended XBRL Tags
The example below depicts a portion of the 2018 income statement of “USA Truck” filed in XBRL format. The “operating revenue” is captured by the Taxonomy tag <us-gaap:Revenues>. This tag repeats three times in the income statement over three different time periods (fiscal years ended 2016, 2017, and 2018). Each instance of the tag is called a fact. To represent the operating expense “Equipment rent” the company uses the extended tag <usak:LeaseAndRentalExpenseEquipment>. The tag is prefixed with the company's ticker (e.g., usak) as opposed to the standard “us-gaap” prefix. This tag also repeats over time and thus represents three different facts.
APPENDIX B
An Illustration of iXBRL
The example below follows Appendix A, illustrating a portion of the 2018 income statement of “USA Truck” filed in iXBRL format. The resulting HTML document contains embedded XBRL tags that are revealed by clicking on the numeric items within the statement. The underlying information includes attributes such as the tag name, value, scale, etc. For example, after clicking the “Operating revenue” (“Equipment rent”) number for the fiscal year ended December 31, 2018, the Revenues (Equipment rent) “Attributes” window appears. The “Attributes” window has four panels, each of which reports additional tag information. While this tag information is also available in traditional XBRL filings, in iXBRL filings it can be accessed directly from the financial statement information presented in the audited HTML report.
APPENDIX C
Obtaining and Using XBRL Data in Academic Research
Appendix C is designed to aid researchers who wish to use XBRL data to answer future empirical questions. We present popular sources for detailed XBRL data that can be extracted and used for future research and provide an overview of how to use XBRL data for textual analysis.48
Custom Computer Programs to Obtain Data Directly from EDGAR
Financial statements prepared in XBRL format are filed with the SEC on an annual and quarterly basis. Each complete filing is comprised of an instance document49 that contains the actual monetary facts, such as the reported amount of revenues and the extension schema that includes the extended tags created by the company. Each filing also includes the following five linkbase files: Label, Reference, Calculations, Definitions, and Presentations. Each linkbase has its own purpose, and a combination of the files is needed to fully analyze XBRL reports.50
Researchers can opt to write their own code to download XBRL files and parse them into a customized database.51 Hence, researchers can configure the database to store data about numeric tags as well as large TextBlocks, enabling textual analysis of financial statement notes and accounting policies.52 This approach requires a significant time commitment and high technical knowledge. To help potential XBRL users, the SEC and other data providers offer more easily accessible data.
Using Custom Computer Programs to Extract XBRL Data for Textual Analysis
XBRL is used to tag the financial statement information at different levels, each enabling researchers to perform textual analysis in varying levels of granularity. First, researchers can extract Level 1 footnote tagging which requires each footnote to be tagged in its entirety as a single text block. Using the Level 1 TextBlocks can aid researchers in identifying and extracting textual information in each footnote. More importantly, this should help researchers perform cross-sectional comparisons because companies should use the same tag name to depict a particular footnote. For example, the tag “us-gaap:FairValueDisclosuresTextBlock” repeats across 73 percent of firms that report fair value (Ahn et al. 2020a).
Level 2 requires the tagging of each significant accounting policy, enabling the examination of accounting policies. For example, Level 2 tagging would enable researchers to extract and examine companies' revenue recognition policies. Level 3 requires that firms separately tag each table or schedule, making it easy to identify and extract tables and their contents such as the tax reconciliation schedule. Level 4 requires firms to tag each individual numerical value that appears in a table, or within a text segment, enabling researchers to accurately separate numeric from textual disclosures. Overall, XBRL enables more accurate textual analysis of disclosures in the notes to the financial statements.
SEC Dataset
The SEC provides datasets of the financial statement and notes that are filed using XBRL. These files are available at https://www.sec.gov/dera/data/financial-statement-and-notes-data-set.html. The SEC provides eight files with various information. The “SUB” file contains information about each filing, which include the company name, unique filing number (accession number), industry membership, and other pertinent information. Other files contain information about each XBRL tag (“TAG”), including information about the tag name, the version of the taxonomy used, whether it is a taxonomy tag, and information on the data type (i.e., monetary, percent, etc.). The “DIM” file contains dimension-related information that can be used to identify additional properties about each tag, such as Levels 1, 2, and 3 fair value as well as segment-related information. The “NUM” and “TXT” files contain information on the value of numeric and textual tags, respectively.53 Other files include information about the rendering process (“REN”) which is the process of making XBRL readable by people. The presentation (“PRE”) file shows how tags are presented in the primary financial statements and the calculation file (“CAL”) provides information on the arithmetical relationships between some tags. Archival researchers should find these files easy to work with as they have already been parsed by the SEC.
Overall, this is a great source of data due to its ease of extraction and interpretation, but it is not without limitations. Researchers intending to use the SEC files for textual analysis should use caution because the SEC limits the field size to 2,048 bytes. Hence, since most TextBlock tags contain larger text, the SEC essentially truncates the data. Therefore, using the SEC provided TextBlock tags for textual analysis is not feasible.
Processed Data
Similar to the mature NLP literature, (e.g., Loughran and McDonald—https://sraf.nd.edu/), some researchers share processed XBRL data. For example, Hoitash and Hoitash (2018) develop a measure of accounting reporting complexity (ARC) based on the count of distinct tags in XBRL filings and found that greater complexity is detrimental to the preparation and audit of the financial reports. Data based on this study is constructed from the SEC database and updated annually. The data can be found at https://www.xbrlresearch.com. The database includes several permutations of the ARC measure for annual and quarterly filings as well as additional measures, such as the number and percent of extended tags, accelerated status of the filers and the filing date and time.
XBRL US
XBRL US is a non-profit organization with the mission to support the implementation of XBRL business reporting standards. The organization provides a database of public filings for XBRL US members, available at: https://xbrl.us/home/use/filings-database/. The database can be linked to a local database installation and queries can be easily answered.
XBRL Data Providers
Several data vendors offer XBRL based data used by public accounting firms, universities, the Department of the Treasury, Morgan Stanley, and the SEC. Calcbench is one such data aggregator that extracts data from XBRL and provides varying sets of information. For example, users can download as-reported XBRL data from financial statements and notes. In addition, similar to traditional data aggregators such as Compustat, Calcbench provides its users with standardized data that can be downloaded and analyzed. Calcbench also allows users to access and download detailed data from financial statement notes that otherwise must be hand-collected.54 Using data aggregators such as Calcbench can allow researchers to use XBRL data without the need to create custom computer programs.55
See Section II for history and a technical overview of XBRL.
Our synthesis of the literature is generally focused on U.S. implications of XBRL reporting, with some discussion of international implications. Globally, more than 100 countries have some form of voluntary or mandatory XBRL reporting (Cohn 2017), with a large concentration of participating countries in Asia, Europe, and South America. Many of these countries, like the United States, make XBRL data publicly available. A list of some countries whose interactive data are made available electronically may be found at https://www.xbrl.org/the-standard/why/ten-countries-with-open-data/.
Each accounting concept, e.g., revenue, is represented by an XBRL tag, e.g., <us-gaap:Revenues>. When available, companies are expected to use standard tags that appear in the XBRL U.S. GAAP Taxonomy maintained by the FASB. When companies deem that no appropriate tag is available to represent a unique company-specific accounting concept, they can extend the taxonomy and create their own tag. We provide an illustration of standard and extended tags in Appendix A.
For example, the name of the tag (e.g., Revenue) as well as other attributes such as scale, period (e.g., December 31st, 2019), and whether the value is monetary, textual, or other, provide contextual information about the disclosed value.
iXBRL will be phased in over a three-year period based on filer status beginning with fiscal periods ending on or after June 15, 2019. Regulators believe that this will lead to a reduction in the burden on preparers, greater preparer autonomy over the presentation of XBRL disclosures, enhancement to the usability of XBRL disclosures, and improved quality. More information on iXBRL is discussed in Section II.
This section is adapted from information available on the xbrl.us (https://xbrl.us/) website and from Hoitash and Hoitash (2018).
Foreign private issuers are required to furnish financial statements in XBRL format for fiscal periods ending on or after December 15, 2017, with no phase-in based on filer status.
See Appendix C, “Obtaining and Using XBRL Data in Academic Research,” for further information on different levels of tagging financial statement footnotes.
IFRS filers are required to file in iXBRL for fiscal periods ending on or after June 15, 2021.
While XBRL requirements currently apply to the 10-K and 10-Q, proponents of the data have called for a mandate to extend to earnings releases, proxy statements, and management commentary (Singh 2018). XBRL requirements have also been considered for the auditors' report; Cohen, Debreceny, Farewell, and Roohani (2014) describe the technical considerations of tagging the auditors' report in XBRL format.
Prior to XBRL reporting, two primary barriers prevented this type of streamlined comparison and analysis. First, it was difficult to accurately extract and identify information from HTML (Hyper Text Markup Language) filings, the other financial statement disclosure format, because these filings were designed only for human consumption. Second, firms used different terminologies and syntax to describe their accounting concepts which prevented computers from being able to consistently identify the same accounting concept across different firms. For example, three firms can use the terms “Operating loss,” “Loss from operations,” and “Income from Operations” to describe the same accounting concepts and all three firms will use the same XBRL tag <us-gaap:OperatingIncomeLoss> to capture that information.
For example, the tag “us-gaap:FairValueDisclosuresTextBlock” captures the entire fair value footnote, the tag “us-gaap_FairValueOfFinancialInstrumentsPolicy” captures the fair value measurement accounting policy, and the tag “us-gaap_ScheduleOfFairValueAssetsAndLiabilitiesMeasuredOnRecurringBasisTableTextBlock” captures the tabular disclosure of assets and liabilities measured at fair value, which is a common table within the fair value footnote.
For example, access to an XBRL based measure of accounting reporting complexity is available at http://www.xbrlresearch.com/accounting-reporting-complexity/. The entire text of frequent financial statement notes (TextBlocks) can be downloaded from http://www.xbrlresearch.com/financial-statement-notes/.
Annual updates include the elimination of unnecessary tags and addition of new tags, including those related to changes in accounting standards.
Tag metadata includes information about the scale of the amount reported, the associated FASB accounting standards codification, and whether the amount is a positive or negative.
For example, since firms are required to report comparative data over a three-year period, typical 10-K filings include at least three instances of the revenue tag with three different values (i.e., facts).
The XBRL system has been updated to allow issuers to associate an extended tag with a taxonomy tag. This practice, referred to as anchoring, provides context for the relationship between an extended tag and a taxonomy tag. This practice is voluntary in the United States and mandatory for iXBRL filings in the European Union.
Journal rankings were identified from the Australian Business Dean's Council, available at https://abdc.edu.au/research/abdc-journal-list/.
The majority of our review was concluded in 2019. Some working papers were published, updated, or added after our initial collection of literature.
We highlight working papers and papers published in top accounting journals and the Journal of Information Systems. Journals included in the “Other” category include: Accounting Horizons; Current Issues in Auditing; Decision Support Systems; International Journal of Accounting Information Systems; International Journal of Disclosure and Governance; Journal of Accounting and Public Policy; Journal of the American Taxation Association; Journal of Business Research; Journal of Emerging Technologies in Accounting; and Journal of Financial and Quantitative Analysis.
Our review also includes publicly available papers that we know to have used XBRL data but were not identified using these key search terms.
Our study focuses on how XBRL influences accounting and auditing. Prior literature reviews on XBRL have examined the role of XBRL in accounting information systems (Gray, Chiu, Liu, and Li 2014; Perdana, Robb, and Rohde 2015; Hutchison, Daigle, and George 2018; Chiu, Liu, Muehlmann, and Baldwin 2019) and enterprise resource planning (Grabski, Leech, and Schmidt 2011).
Some papers are included in multiple categories.
The number of identified papers using numerical or footnote (textual) XBRL data is likely understated. Some papers may unknowingly use XBRL data provided by third-party data providers.
In an international setting, survey evidence suggests that German finance managers have greater incentives to adopt XBRL than IT managers because of greater cost considerations and more direct impact on financial reporting (Pinsker and Felden 2016). Examining countries with both voluntary and mandatory adoption of XBRL, Abdolmohammadi, DeSimone, Hsieh, and Wang (2017) find that the internal audit function is more likely to be involved in XBRL implementation for large companies in common law countries than for small companies in civil law countries, likely driven by stronger corporate governance, greater economic resources, and stronger investor protection available to companies in common law countries.
Janvrin and Mascha (2014) conduct a field investigation about the financial close process and find that XBRL tagging does not cause delays.
Liu et al. (2014a) document an increase in cost of capital in China during the early mandatory adoption period, suggesting an increase in transaction costs.
In the international context, XBRL appears to reduce information asymmetry in Europe (Liu, Luo, and Wang 2017; Kaya and Pronobis 2016), Korea (Yoon, Zo, and Ciganek 2011), and China (Chen, Guo, and Tong 2017a). Wang and Seng (2014) suggest that Chinese investors do not use XBRL data but find an increase in foreign institutional investors of Chinese firms post XBRL adoption, likely due to reduced information asymmetry and processing costs.
Another stream of literature considers how firms respond to the reduction in investors' information processing costs post-XBRL mandate. XBRL reporting, or the change in shareholders' information processing costs, results in changes in tax avoidance strategies (Chen, Hong, Kim, and Ryou 2017b) and increases in footnote disclosures (Blankespoor 2019).
While data use concerns declined during the mandatory filing period, this issue may still be of concern in studies examining the years immediately following mandatory adoption because data on the actual use of XBRL is not available.
Examining the SEC's goal of leveling the playing field between investor types, Hodge, Kennedy, and Maines (2004) conduct an experiment and find that many nonprofessional investors fail to utilize XBRL filings, but that those who do are better able to acquire and integrate information. Offering contradicting evidence, Janvrin, Pinsker, and Mascha (2013) find that most investors (66 percent) use XBRL based on a perceived increase in task efficiency.
Bhattacharya, Cho, and Kim (2018) find similar evidence of a leveling of the playing field, measured using trading responsiveness, between large and small institutional investors.
This is consistent with the experimental evidence offered by Pinsker and Wheeler (2009), who find that non-professional investors perceive XBRL-formatted documents to be more efficient and effective than other formats. Arnold, Bedard, Phillips, and Sutton (2012) offer evidence that nonprofessional investors' search strategies are improved when narrative disclosures are tagged in XBRL format.
However, Brown, Cohen, and Huffman (2019) find that the positive association between accounting reporting complexity and non-GAAP disclosures is driven by standard taxonomy elements rather than firm-specific elements (extended tags).
Management's involvement in the detailed tagging process is unclear. Many preparers rely on the help of software provided by vendors or fully outsource the XBRL conversion process. Additionally, using XBRL to obfuscate financial reporting would require a sophisticated understanding of XBRL that management may not possess.
The Figure begins in fiscal year 2012 because this was the first year of mandatory detailed tagging of the financial statements and footnotes by all filers.
However, Johnston (2020) documents that these results apply only to extensions of information disclosed in the financial statement footnotes rather than extended tags of information recognized on the face of the financial statements. Li and Nwaeze (2015) find that extensions exacerbate information asymmetry in early years of firms' XBRL adoption, but improve firms' information environments in later years.
Kim, Kim, and Lim (2019a) find that the use of standardized XBRL tags constrains earnings management.
Smith, Zhang, and Kipp (2019) measure ARC based on the number, percent, and ratio of customized XBRL tags, also finding that ARC influences auditors' likelihood to issue a material weakness.
The association with misstatements applies only to the percentage of extended tags.
We are aware of two Big 6 accounting firms that use the services of Calcbench, an XBRL data provider. While other firms may also subscribe, or obtain XBRL data from other sources, this information is not publicly available.
If users cannot trust the accuracy of the data, they are less likely to use it.
Frameworks for how to audit XBRL are proposed by Srivastava and Kogan (2010) and Boritz and No (2016).
Of the countries requiring financial statements filed in XBRL, a limited number have some form of assurance requirement. As of January 1, 2018, The Netherlands was the first country to require assurance of XBRL filings by the external auditor. In November 2019, the Committee of European Auditing Oversight Bodies mandated and offered guidance on the audit of iXBRL filings. If XBRL tags are materially misstated, auditors should express a qualified or adverse opinion on XBRL compliance. Starting with fiscal years beginning January 1, 2020, for companies trading on regulated markets in the European Union, the European Single Electronic Format Regulation requires the auditor to opine on financial statement compliance with XBRL tagging requirements. In India, an external accountant is required to certify that the XBRL filings are complete, accurate, and fairly represent the financial statements. However, this certification is not equivalent to a complete audit of the filings (Farewell et al. 2017).
An example of the metadata available in iXBRL filings is provided in Appendix B.
While assurance is not required, anecdotal evidence confirms that some companies choose to involve their external auditor in the review of XBRL filings.
Additional research examining the adoption of iXBRL may consider the role of individuals responsible for making this decision (Pinsker and Felden 2016).
We suggest that researchers using XBRL data begin the sample period in 2012, the first year in which most companies were required to tag both the financial statements and accompanying footnotes. Data prior to 2012 is thus biased toward larger companies who were required to fully comply with the XBRL mandate at an earlier stage of the phase-in, or includes companies who were only required to tag the face of the financial statements.
With the introduction of iXBRL, firms no longer provide an instance document. This document is now replaced with an HTML file that embeds the XBRL tags.
The Label linkbase provides human readable strings for each accounting concept. The Reference linkbase provides reference to authoritative accounting standards (e.g., the FASB codification). The Calculation linkbase presents the mathematical relationships between certain tags. The Presentation linkbase instructs how to present the different concepts and organize them into financial statements and notes. The Definition linkbase associates non-calculation relationships between concepts.
Links to all XBRL filings can be found at https://www.sec.gov/Archives/edgar/monthly/
A significant limitation of XBRL filings obtained from any source is that identifying the exact location of numeric XBRL facts (i.e., in which financial statement, note, or table) is a difficult task because many tags repeat across multiple statements and notes. While it is possible to partially overcome this limitation with extensive programming and the use of heuristics, this is a major limitation of current XBRL filings irrespective of the data source. This issue may be resolved with the introduction of iXBRL.
For example, information on a particular segment typically appears as a dimension.
For example, while in prior years researchers had to manually collect purchase price allocation of M&As, using XBRL data, Calcbench automates the collection of such data.
Another popular data vendor used by academic researchers is XBRL Cloud, available at www.xbrlcloud.com.