Auditor benchmarking of client disclosures

For example, the Public Company Accounting Oversight Board (PCAOB) AS 2810, paragraph 31, states that auditors are responsible for “the form, arrangement and content of the financial statements (including the accompanying notes), encompassing matters such as the terminology used, the amount of detail given, the classification of items in the statements, and the bases of amounts set forth.”

In our first hypothesis, we discuss that disclosure benchmarking could be either positively or negatively associated with the existence of authoritative audit guidance, which either could help resolve reporting uncertainty or exist in areas with substantial reporting uncertainty.

As an example of an auditor who had a clear sense of whom to use as benchmarks, one respondent reported the following information about working in the aerospace industry: “I primarily work in aerospace and defense and my client, a large accelerated filer, did not early adopt ASC 606. They’ll have to disclose more granular revenue detail in a footnote for the first time in Q1, so I’ve been looking over the disclosures of [aerospace industry peer firm #1] and [aerospace industry peer firm #2] to see how they disaggregated revenue in their recent 10-Ks since [both peer firms] early adopted.”

Regulatory filings (e.g., Forms 10-K and 10-Q) are legal documents that include information beyond audited or reviewed financial statements. Because auditors are required to read and consider information that accompanies audited or reviewed financial statements for material consistency, we speak generically about the external auditor auditing or reviewing SEC Form 10-K and 10-Q information.

Based on our collective audit experience and on discussions with auditors, we expect that auditors will not prefer less disaggregated disclosure, even when less disaggregated disclosure is consistent with the disaggregation practices of benchmark firms.

For example, a 2010 FASB Exposure Draft refers to disaggregated information as one of two “Core Principles of Financial Statement Presentation” (FASB 2010).

“EDGAR” is an acronym for Electronic Data Gathering, Analysis, and Retrieval. The SEC describes and provides access to EDGAR at the following location: https://​www.​sec.​gov/​edgar/​searchedgar/​companysearch.​html.

See, for example, https://​whois.​arin.​net/​rest/​net/​NET-167-219-0-0-1.

Note that IP addresses themselves can be anonymized via the usage of virtual private networks (VPNs). As a result of VPN usage, a single IP address can reflect the activity of more than one user. VPN usage is common at Big Four firms, so a single IP address often supports multiple auditors via the VPN. For this reason, we include all downloads for all filings, rather than removing duplicates based on the IP address, under the assumption that duplicate observations are likely legitimate, because they represent multiple users that fall under a single IP address.

In an untabulated analysis, we find evidence that the non-Big Four audit firms (e.g., BDO and Moss Adams) also access a significant number of SEC filings; however, relative to the Big Four, they access a much smaller number. Therefore, like many other studies, we focus only on the Big Four audit firms and their clients.

We focus on same-industry downloads for several reasons. First, there are financial reporting norms and audit conventions that directly tie to the client’s industry. Companies operating in the same industry are more likely to have similar transactions and face similar reporting conditions, relative to companies in other industries. Second, and similarly, companies in the same industry often face additional regulatory requirements that apply widely across the industry (e.g., airlines or oil and gas) but do not apply to other industries. Third, companies in the same industry likely face similar supply and demand conditions and are subject to similar microeconomic forces (e.g., factor and labor inputs, technologies) and macroeconomic forces (e.g., booms and busts or cyclicality). Further, practicing auditors indicated that, when seeking out SEC filings for benchmarking purposes, they are most likely to seek those from companies within the client’s industry. The survey evidence presented in Section 2.1 indicates that the most frequently mentioned response to the question “Who do you look for when seeking out filings?” was “same industry” (see Table 1 for the questions and response rates).

Our measure captures the auditor’s acquisition of non-audit clients’ financial statements but does not capture their acquisition of its own clients’ financial statements. We make this empirical choice in part because an auditor acquisition of its own client’s filings can be driven by audit motives that are outside the scope of auditor benchmarking (e.g., archiving, compliance, inter-office communications, etc.). Furthermore, at the construct level, we are interested in how auditors learn from the disclosure practices of other auditors, and, because it captures auditors’ usage of non-client financials, our measure is well-suited to capture that construct.

These guides serve as a proxy for authoritative guidance in the industry. The industries that have these guides are airlines, construction, investment management, broker-dealers, gaming, oil and gas, banks and financial institutions, healthcare, and property and casualty insurance. Because these are industry effects, we do not include industry fixed effects in estimating model (1) below.

Specifically, we employ the predicted litigation risk measure based on Model 3a of Kim and Skinner (2012). We re-estimate their model parameters over our sample period for our sample firms.

For example, Oracle, Microsoft, and Workday each have a four-digit SIC of 7372. Thus we include matched-pair fixed effects for the Oracle-Microsoft pair, the Microsoft-Workday pair, and the Oracle-Workday pair, in which the auditor of the former is searching for the SEC filings of the latter.

In fact, one of the survey respondents from Table 1 indicated that, during the course of disclosure benchmarking, he or she would seek out information on how other firms and their auditors had disclosed material weaknesses.

As discussed by Chen et al. (2015, p. 1019), the measure of disaggregation excludes cash flow items, MD&A disclosure, per share Compustat data items, and calculated items within Compustat, which are not disclosures made by the company. The measure excludes cash flow statement items, because little variation exists in the disclosure of cash flow statement information. The exclusion of MD&A disclosure from the annual financial reports appears reasonable, because the MD&A is a discussion of the results of operations (income statement) for the given fiscal year and the financial condition (balance sheet) as of year-end. We are grateful to the authors of Chen et al. (2015) for sharing their data.

For this test, our sample size declines, due to the requirement that the auditor access at least one benchmark firm annual report during the filing review period and due to cases where a particular set of footnotes was not available in the dataset of financial statement footnotes we employed.

Following Lang and Stice-Lawrence (2015), we calculate the cosine similarity of the financial footnotes between client and benchmark firms after deleting words not found in the Loughran and McDonald (2014) English business dictionary: https://​sraf.​nd.​edu.

We are grateful to Jeff McMullin for graciously providing us with cosine similarity scores, using his proprietary financial statement footnote data.

The positive coefficients on DISCLOSURE BENCHMARKING in Columns 1 and 2 are consistent with a positive relationship between disclosure benchmarking and client financial statement disaggregation. The positive coefficient also allows for the possibility that decreases in disclosure benchmarking are associated with decreases in client financial statement disaggregation. In untabulated analyses, we measure increases in disclosure benchmarking by replacing DISCLOSURE BENCHMARKING with an indicator variable equal to one when disclosure benchmarking increases over the prior year and zero otherwise. We find a positive and significant coefficient on the indicator variable, consistent with H2 and consistent directional increases in disaggregation subsequent to benchmarking. In a separate untabulated test, we restrict the sample to years when DISCLOSURE BENCHMARKING changed year over year. The results of this test also remain consistent with those reported in Table 6.

An interesting possibility is that, absent frictions, firm disclosure and disaggregation will increase almost infinitely over time (i.e., auditors will see that other companies provide increasingly disaggregated financials and will request increased disaggregation of their own clients, which will be seen by other auditors, who will request even further disaggregation). Indeed, Chen et al. (2015, p. 1032) report that financial statement disaggregation is increasing year over year. However, auditor benchmarking and disclosure changes by management are likely associated with significant informational frictions, such as costs for information acquisition and processing and compliance. These frictions will likely hinder perfectly revealing, or fully disaggregated, disclosure.

We note that relatively few coefficients are statistically significant in Columns 3 and 4 of Table 6. Further, the model exhibits low explanatory power, particularly in comparison to the explained variation in the disaggregation models (Columns 1 and 2). Research that uses text-based comparability measures as dependent variables also exhibits fairly low explanatory power (< 20%), suggesting these types of dependent variables are somewhat noisy (e.g., Brown and Tucker 2011; Peterson et al. 2015; Lang and Stice-Lawrence 2015). We acknowledge the low explanatory power and the number of insignificant independent variables is a potential limitation with this analysis.