The Impact of Trade Secrecy Protection on Audit Pricing

Abstract

Because auditors have access to corporate information, a firm’s decision to protect material trade secrets should, in principle, not influence audit effort. We analyze the effects of trade secrecy protection on the audit fees, documenting that firms with redacted information pay significantly higher fees than those that do not redact information. In cross-sectional tests, we further document that the relationship between redaction and audit fees is significantly influenced by both auditor and client characteristics. Consistent with the premise that redaction increases the complexity of the audit—particularly if auditors view redacted disclosures as indicators of potential business or litigation risk—the regression results indicate that the main effect is moderated by auditor factors such as specialization, tenure, and quality, as well as client factors like existing relationships, bargaining power, and reporting quality. These insights contribute to ongoing discussions in audit policy by illustrating how confidential disclosure practices affect audit effort and costs. Overall, our results inform policymakers seeking to reconcile firms’ proprietary information protection with public interest in transparent and credible financial reporting.

1. Introduction

Protecting trade secrets is crucial in maintaining a firm’s competitive position, yet extant reporting regulations frequently compel firms to disclose proprietary information. To balance the interests of the public and those of the firms, the SEC in the U.S. permits companies to redact information from material contracts in a firm’s regulatory filings, provided that the information is both ‘proprietary and immaterial’ to investors (Thompson, 2022; Thompson et al., 2023). One line of research explores firms’ incentives for redaction, showing that both concerns over proprietary costs and agency-related motives (e.g., withholding negative information) drive the decision to redact (Verrecchia & Weber, 2006; Thompson et al., 2023). Another body of literature investigates the consequences of redaction, finding that while redacting firms generally exhibit better accounting performance (Lee, 2022), they also face increased information asymmetry in capital markets (Boone et al., 2016; Thompson, 2022).

This study addresses the unexplored question of how firms’ decisions to redact material information—intended to protect trade secrets—affect audit pricing and engagement risk. Although prior research has examined the capital market and regulatory consequences of redaction, less is known about its implications for auditors who must evaluate the credibility and completeness of clients’ financial reporting. Our objective is to determine whether and how redaction alters auditors’ perceived risk and pricing of their services.

Prior literature suggests that audit fees increase with auditors’ engagement risk, which can take three forms: (1) a client’s business risk, reflecting concerns about profitability and survival; (2) the risk of material misstatements; and (3) the auditor’s business risk, involving reputational harm and potential litigation and arising from undetected misstatements in financial reports (Bell et al., 2008; DeFond et al., 2016). Redaction of proprietary and other information has significant consequences for the firm’s information disclosure environment: it can increase client business risk to the extent that managers use redaction to conceal bad news (Bao et al., 2022), it can obfuscate the firm’s information disclosures (Verrecchia & Weber, 2006; Boone et al., 2016; Thompson, 2022), and it can elevate the auditor’s business risk if regulators such as the SEC and the Internal Revenue Service (IRS) increase monitoring of the redacting firm (Gietzmann & Pettinicchio, 2013; Kolev et al., 2023; Bao et al., 2023), all of which can increase the auditor’s litigation and reputational risk. While auditors have access to their clients’ information (i.e., can access redacted information with little effort), they may charge higher fees for redacting firms if they perceive that redacted disclosure increases any aspect of their engagement risk. Therefore, we hypothesize and test whether firms that redact disclosures face higher audit fees compared to those that do not.

We test our hypothesis based on 73,517 firm-year observations from 2000 to 2019. We identify a firm’s redaction status using its 10-Q and 10-K filings following the procedures in G. Chen et al. (2022) and show that redacting firms pay significantly higher fees to their auditors than those without redaction. Additional analyses using a time-series model and an instrumental variable approach based on customers’ trade secrecy both confirm that the higher fees incurred by redacting firms are not driven by time-invariant characteristics but reflect auditors’ response to increased engagement risk. We further categorize the redacted information into five main groups related to executive compensation, financial contracts, supplier/customer contracts, intangible contracts, and unspecified items and find the higher audit fees are most evident for redactions related to business dealings with suppliers and customers, and for intangible assets such as license agreements, software, and patents.

Finally, we perform several cross-sectional analyses to analyze potential mechanisms driving the increase in audit fees among firms that engage in redaction. Specifically, we show the fee increase is mitigated for companies audited by Big 4 firms, auditors with a longer tenure, and specialist auditors. Additionally, the fee increase is less pronounced for clients with longer auditor relationships, larger clients with greater bargaining power, and those producing higher-quality financial reports. These results align with the notion that Big 4 auditors, by leveraging their extensive client knowledge, larger resource bases, accumulated experience, wider networks, and internal expertise in areas such as valuation, tax, and legal matters (Francis & Yu, 2009; Sundgren & Svanström, 2013), are better equipped to manage various audit engagement risks (DeFond & Zhang, 2014; X. Chen et al., 2020), and that client-specific knowledge and higher financial reporting quality help auditors reduce potential business or litigation risks associated with changing disclosure (Heinle et al., 2023; Barth et al., 2023).

Our study provides several related contributions to the audit and disclosure literature. First, we contribute to the redaction literature and, more broadly, the literature on economic consequences of trade secrecy protection. While previous studies offer insights into the economic consequences of redaction, they primarily focus on its effects on investors (Bao et al., 2022; Verrecchia & Weber, 2006; Boone et al., 2016; Thompson, 2022). The evidence beyond the stock market remains limited, except for Kolev et al. (2023), who examine how firms’ management of proprietary information interacts with SEC monitoring, manifested in a higher incidence of comment letters and a longer resolution process among redacting firms. We extend this research by presenting evidence of a previously underexplored effect on auditors. Second, our study extends prior research on the effects of corporate disclosures on audit fees. Existing studies have primarily focused on the response of audit fees to various types of disclosures, including management forecasts (Krishnan et al., 2012) and risk disclosures in 10-K filings (Yang et al., 2018), or properties of regulatory filings, such as litigious tone (Malik et al., 2022) and readability (Blanco et al., 2021). Despite the importance of sensitive information in shaping corporate disclosure policies (Verrecchia, 1983), whether and how firms’ nondisclosure choice is related to audit fees remains relatively unexplored. Overall, our study is novel in linking trade secrecy protection to audit outcomes, thereby integrating the literature on proprietary disclosure and audit economics. By documenting that redacting firms pay higher audit fees and that this effect depends on auditor and client characteristics, we contribute new evidence on how information opacity shapes audit policy, assurance costs, and ultimately, financial transparency.

Our findings carry important implications relevant for audit policy and financial transparency. By showing that redaction of proprietary information increases audit fees through higher perceived engagement risk, we highlight how disclosure practices influence the cost and effectiveness of audit oversight. Overall, these results suggest that audit regulators and standard setters should consider how confidentiality provisions interact with transparency objectives in ensuring credible financial reporting.

2. Prior Literature and Hypothesis Development

Firms are required to provide disclosures to protect the interests of their investors. However, such disclosures could potentially undermine their competitive edge if the revealed information proves to be advantageous to their competitors. In an effort to mitigate the potential competitive harm stemming from mandatory disclosure requirements, the SEC allows firms to file confidential treatment requests (CTRs), enabling them to withhold proprietary information that would otherwise be publicly disclosed (Kolev et al., 2023). The majority of CTRs are filed to protect proprietary information contained in contracts (e.g., material contracts with suppliers and customers or intangible assets such as license agreements, software, and patents) that would otherwise be included in SEC filings such as 10-Q and 10-K filings (Fei, 2022).

The process for filing a CTR is outlined in the Staff Legal Bulletin No. 1.1 Specifically, the company needs to file the application via the SEC’s EDGAR system, and specify (1) the content to be redacted; (2) the immateriality of the redacted information; (3) how the disclosure of such information would lead to substantial competitive harm; (4) the desired duration of the CTR; and (5) the original version of the exhibit without redaction. The SEC reserves the right to approve, deny, or request supplementary details regarding the CTR. If the SEC approves the CTR, it grants a confidential treatment order specifying the duration of confidentiality of the redacted information, which is usually the contract’s minimum term or 10 years, whichever is shorter. Should the CTR be partially approved or rejected, the firm needs to submit an amended version of the original filing (Fei, 2022).

2.1. Literature Review

An emerging body of research explores the factors influencing information redaction and its consequences. Regarding the determinants, prior studies suggest that firms redact information primarily due to proprietary cost concerns (Cao et al., 2018; Bourveau et al., 2020; Glaeser, 2018; Saavedra, 2023; G. Chen et al., 2022; Floros et al., 2025; Zheng, 2023) or to conceal unfavorable news (Bao et al., 2022). In addition to proprietary cost and agency concerns, firms may also redact to avoid IRS scrutiny (Bao et al., 2023) or to strengthen their bargaining position with labor unions by suppressing wage demands (Yoo, 2023).

In terms of consequences, while the market generally responds positively to the voluntary nondisclosure of proprietary information (Lee, 2022), there is evidence of a negative impact on the firm’s information environment and oversight. For instance, redacting firms face increased adverse selection (Verrecchia & Weber, 2006), heightened information asymmetry before initial public offerings (Boone et al., 2016), slower stock price discovery, more insider trading (Thompson, 2022), and less accurate and more dispersed analysts’ earnings forecasts (Fei, 2022). To counter these adverse effects, redacting firms have voluntary increased disclosure of non-proprietary information, which has led to a net decline in corporate transparency and a more opaque information environment (Glaeser, 2018; Barth et al., 2023; Heinle et al., 2023), and increased the likelihood of redacting firms becoming takeover targets (Griffin et al., 2022) and falling victim to cybersecurity breaches (Hughes et al., 2023). Others, such as Kolev et al. (2023) argue that redacting proprietary information heightens information uncertainty, and when combined with the interplay between CTR oversight and periodic filings, it can attract regulatory attention. This heightened scrutiny increases the likelihood that SEC staff will detect deficiencies in a firm’s regulatory filings. Consistent with their predictions, they find that when redacting firms receive comment letters, the resolution process takes longer and requires more SEC resources, suggesting that such letters reflect more intensive regulatory oversight and a greater focus on identifying financial reporting deficiencies.

The body of research examining the relation between corporate disclosures and audit fees is vast and shows that auditors charge higher fees for various types of disclosures that are likely to entail higher reputational, earnings management, or litigation risk, in settings that involve annual or quarterly management earnings forecast (Krishnan et al., 2012), pro forma earnings (L. Chen et al., 2012), risk factor disclosures in 10-Ks (Yang et al., 2018), goodwill-related disclosures (V. Chen et al., 2019), and Level 3 fair value measurement (Hong & Hwang, 2018). Firms also incur significant proprietary costs of disclosure (and higher audit fees) when facing a high degree of technological industry pressure (Cao et al., 2018), when they reach a certain size threshold (Bernard et al., 2018), and when they are subject to whistleblowing allegations (Kuang et al., 2021). Audit fees are also significantly associated with the properties of financial reporting, including accrual quality (Cho et al., 2017), litigious tone in annual reports (Malik et al., 2022), and 10-K filing readability (Abernathy et al., 2019; Blanco et al., 2021).

These results suggest that auditors may intensify their efforts in order to mitigate litigation and reputational risks or charge a risk premium for the higher anticipated litigation costs tied to disclosures. A related body of research further explores the connections between firm opacity, disclosure quality, and audit outcomes.2 Danielsen et al. (2007), for example, suggest that auditors’ awareness of a firm’s opacity can lead to higher audit fees, as firms may pay premium fees to purchase reputational capital and enhance markets’ perception of transparency. Alternatively, as Danielsen et al. (2007) find, higher audit fees may reflect the increased risk faced by auditors, when they evaluate firms with opaque disclosures. Subsequent studies have introduced new methods for estimating firm opacity (e.g., Corwin & Schultz, 2012) and demonstrate that redaction may heighten the opacity of financial reports, as managers tend to withhold more proprietary information than is necessary (Thompson et al., 2023). Although existing studies highlight various consequences, including reduced transparency, improved disclosure, increased information asymmetry, and shifts in regulatory oversight following redaction, we provide empirical evidence of how these factors directly influence auditor engagement risk and the corresponding audit fees.

2.2. Hypothesis Development

It is a priori not obvious whether redacted disclosure is likely to impact audit fees. First, auditors have direct access to the management and can obtain private information to assess audit risks whenever necessary (DeFond & Zhang, 2014; Hribar et al., 2014). Second, auditors collect and evaluate all client-specific information (both quantitative and qualitative) when determining audit fees (Simunic, 1980; Abernathy et al., 2019; Brazel et al., 2009). To the extent that firms increase voluntary disclosures after redaction (Glaeser, 2018; Heinle et al., 2023; Barth et al., 2023), the auditors’ overall engagement risk may not change significantly.

We suggest that the causal mechanism linking redaction to higher audit fees is primarily driven by the auditor’s perception of increased overall engagement risk. Although auditors can access redacted information with little effort, the act of redaction itself can send a signal to the auditor that can increase their perceived risk, uncertainty, and, consequently, their fees.3 Specifically, redaction signals to auditors that management has intentionally withheld information from public stakeholders, citing confidentiality or competitive harm. This act raises a red flag for auditors because it may indicate a broader pattern of reluctance to disclose potentially sensitive or unfavorable information, which can increase the auditor’s engagement risk. Prior research suggests that engagement risk involves three forms, including: client’s and auditor’s business risk, and the risk of material misstatement. We argue that all three forms of engagement risk can be elevated by redaction. First, redaction can increase a client’s business risk if managers are using it to conceal bad news. This obfuscates the firm’s information environment, leading to increased information asymmetry. Second, the act of redaction, and the potential for a broader strategy of withholding information, may signal an increased risk of misreporting and disclosure deficiencies. Third, redaction can increase the pressure on auditors to identify potential misstatements, as a higher regulatory oversight could lead to litigation and reputational damage.

Prior evidence on disclosure and regulatory oversight provides supporting evidence and suggests that redaction may be associated with greater client engagement risk, as it could signal a heightened risk of material misstatement or indicate that managers are strategically concealing unfavorable information (Bao et al., 2022). For example, Kolev et al. (2023) argue that redaction is not merely a decision tied to specific contracts but part of a broader strategy to withhold the information required by disclosure regulations. Boone et al. (2016) find that IPO firms redacting information from one or more material contracts experience greater underpricing and stronger post-IPO financial performance, suggesting that redacting material information prior to the IPO heightens information asymmetry between investors and firms. Similarly, Gietzmann and Pettinicchio (2013) show that the issuance of an SEC comment letter indicates that the firm has come to the attention of the oversight board, causing auditors to fundamentally revise their assessment of the client-specific risk and raise their fees accordingly. Finally, Kolev et al. (2023) and Bao et al. (2023) show that redaction draws a greater attention from regulators such as the SEC and IRS, thus increasing the pressure on auditors to identify potential material misstatements. Drawing on the arguments outlined above, we formulate our first hypothesis (presented in the alternative form)4:

H1. 

Redacting firms pay higher audit fees than nonredacting firms.

We conduct several cross-sectional tests regarding both the supply (auditors’) and demand (clients’) sides of audit services. From the supply side, we test whether the increased audit fees vary with factors such as auditor status, audit tenure, auditor specialization, and the existence of a prior client relationship. First, Big 4 auditors are particularly sensitive to litigation and reputational risks, and are therefore more likely to exert extra effort for redacting firms to mitigate these risks. However, Big 4 auditors, with their extensive knowledge of client operations, larger resource pools, accumulated experience, broad networks, and internal support from specialists in valuation, tax, and legal matters (Francis & Yu, 2009; Sundgren & Svanström, 2013; Bills et al., 2020), are also expected to respond differently to the heightened client engagement risk associated with redacting firms. For example, X. Chen et al. (2020) suggest that Big 4 firms not only have high-quality personnel and stronger monitoring incentives relative to non-Big 4 firms, but also possess more resources that their audit teams can leverage. Other research, such as Hrazdil et al. (2024), shows that Big 4 firms provide superior audit quality due to their experience, expansive peer networks, and enhanced in-house expertise for detecting material misstatements, while Zhang et al. (2018) document that non-Big 4 auditors adjust their fees asymmetrically in response to economic policy uncertainty, whereas Big 4 auditors do so symmetrically. Given their greater resources, we expect Big 4 firms to manage perceived engagement risk from redacted information more effectively, resulting in smaller fee adjustments compared to non-Big 4 auditors. Second, we make similar predictions regarding audit tenure, auditor specialization, and auditors with greater familiarity with their clients, as those with longer relationships and repeated interactions are more likely to have a deeper perception of the client’s business environment, thereby reducing information asymmetry (Karolyi, 2018). Given the opposing views, we formulate the following hypothesis as nondirectional.

H2. 

The positive relation between redaction and audit fees varies depending on auditor characteristics, including firm affiliation, longer tenure, greater industry specialization, and a prior client relationship.

From the demand side, to assess whether higher audit fees are driven by auditors’ need to compensate for increased effort or by the redacting firm’s demand for higher-quality assurance (DeFond & Zhang, 2014), we examine whether the fee increase is different for clients with greater bargaining power and higher reporting quality. First, clients with stronger bargaining power, due to their size, importance to the auditor, or access to alternative auditors, may be better positioned to negotiate lower audit fees (Huang et al., 2007; Choi et al., 2010). Second, while redaction increases information asymmetry and perceived audit risk, high financial reporting quality can mitigate these concerns by enhancing the credibility and transparency of financial disclosures (Francis et al., 2005; Fung et al., 2017). As a consequence, auditors may perceive less incremental risk from redactions among firms with high quality financial reporting, leading to a weaker association between redaction and audit fees. This suggests that the positive effect of redaction on audit fees is attenuated when financial reporting quality is high.

If redacting firms are seeking enhanced assurance from auditors, audit fees should not systematically vary with the client’s bargaining power or reporting quality. However, if auditors are raising fees due to the additional effort required for redacting firms, clients with greater bargaining power and/or higher reporting quality should be able to leverage these advantages to minimize the increased assurance costs. Accordingly, we propose our final hypothesis in its alternative form.

H3. 

The positive relation between redaction and audit fees is weaker for firms with greater bargaining power and higher financial reporting quality.

3. Sample and Research Methods

3.1. Sample Construction

Following G. Chen et al. (2022), we identify companies with redactions using their 10-Q and 10-K filings. We first clean out XBRL and HTML tags, images, and embedded PDFs, and then search for keywords such as “confidential treatment”, “confidential portion”, “rule 406”, “rule 24b2”, or “redacted” within 20 words of ‘agreement’ or ‘exhibit’, or ‘confidential information’ in combination with variations in star marks (G. Chen et al., 2022). Our sample period begins in 20005, the year when redactions became relatively popular, and ends in 2019 due to changes in the regulatory environment. Specifically, the FAST Act Modernization and Simplification of Regulation S-K came into effect in April 2019 with the aim to reduce the SEC’s oversight of redacted contracts, whereby companies no longer need to submit formal CTRs beforehand (Kolev et al., 2023). We then match our main sample with audit-related data from the Audit Analytics and firms’ financial statement data from the Compustat datasets and estimate various controls.

Our broad sample contains 183,285 firm-years (representing 19,515 unique firms) from Audit Analytics for the period 2000 to 2019. First, we exclude 56,668 firm-year observations due to missing data required for calculating control variables or due to negative values for assets, sales, or equity. Next, we remove firms (31,751 firm-year observations) in regulated industries, including utilities (with SIC codes 4900–4999), financial institutions (with SIC codes 6000–6999), and public administration (with SIC codes 9000–9999). Finally, we omit 21,214 firm-year observations with no machine-readable 10-K files in the SEC’s EDGAR archive. These reductions result in a sample of 73,517 firm-year observations across 8670 unique firms, with 5807 observations involving at least one redaction in that year’s filings.

When we breakdown the sample by industry and year, we identify over 50% of the observations that come from the manufacturing sector, which also has the highest rate of redacted filings at 9.75%. About 30% of the observations are from the services industry, with 6.40% of those involving redactions. Finally, the percentage of firms submitting redacted reports increased during the initial phase of the sample period, peaked between 2010 and 2012, and then gradually declined.

3.2. Research Methods

To evaluate our hypothesis, we estimate the following model:

Audit Fee_i,t = β₀ + β₁ ∗ Redaction_i,t + CONTROL_i,t + Firm FE + Year FE + ε_i,t,

(1)

where Audit Fee is the log of the audit fees. Our independent variable is Redaction, a dummy variable, which equals 1 if the firm’s 10-Q or 10-K filings have at least one redaction in a particular year and 0 otherwise. If our hypothesis holds, we expect β₁ to be significantly positive.

Following Hsieh et al. (2020), we incorporate several client and auditor characteristics that have been found in prior literature to influence audit fees. Specifically, we control for Audit Tenure (the log of the number of consecutive years that the client has retained its current auditor), Size (the log of the firm’s total assets), Acc_Filer (equals 1 if the firm is an accelerated filer and 0 otherwise), Leverage (the firm’s leverage ratio; the sum of current liabilities and long-term debt, divided by total assets), Liquidity (the liquidity ratio; the sum of accounts receivables and inventories, scaled by total assets), GC (the going concern indicator; equals 1 if the audit report expresses substantial doubt about the firm’s ability to continue and 0 otherwise), Restatement (equals 1 if the firm announces a restatement in a particular year and 0 otherwise), Loss (equals 1 if a client firm has net loss in a year and 0 otherwise), ROA (return on assets; net income scaled by average total assets), Cash Flow (the ratio of cash flow from operations, scaled by total assets), Foreign (equals 1 if a firm reports any foreign income and 0 otherwise), Segments (the log of the total number of business segments), Report Lag (the log of the number of days between the fiscal year-end and the audit report date), and Big 4 (equals 1 if the auditor firm is one of the Big 4 auditors and 0 otherwise). Appendix A (

Table A1. Variable definition.

Variable	Definition	Data Source
Audit Fee	The log of audit fees.	Audit Analytics
Redaction	Equals 1 if the firm’s 10-Q or 10-K filings have at least one redaction in a particular year and 0 otherwise.	SEC Edgar
# Redaction	The number of redaction-related keywords in 10-Q and 10-K filings (following G. Chen et al., 2022).	SEC Edgar
Post Redaction	Equals 1 if the present year is the year of or is after the year of the first redaction and 0 otherwise.	SEC Edgar
Before_2, Before_1	Before_2 (Before_1) equals 1 if the present year is two (one) year(s) before the first redaction and 0 otherwise.	SEC Edgar
Current	Equals 1 if the present year is the year when the first redaction occurs and 0 otherwise.	SEC Edgar
After_1, After_2+	After_1 (After_2+) equals 1 if the present year is one year (two or more years) after the first redaction and 0 otherwise.	SEC Edgar
Audit Tenure	The log of the number of consecutive years for which the client has retained its current auditor.	Audit Analytics
Size	The log of client firm’s total assets (in millions of dollars).	Compustat
Acc_Filer	Equals 1 if the firm is an accelerated filer and 0 otherwise.	Audit Analytics
Leverage	The leverage ratio; the sum of current liabilities and long-term debt divided by total assets (item (DLT+DLTT)/AT in Compustat).	Compustat
Liquidity	The liquidity ratio; the sum of accounts receivable and inventories, scaled by total assets.	Compustat
GC	Equals 1 if the audit report expresses substantial doubt about the client’s ability to continue as a going concern and 0 otherwise.	Compustat
Restatement	Equals 1 if a client firm announces a restatement in the fiscal year and 0 otherwise.	Audit Analytics
Loss	Equals 1 if a client firm has a net loss in the current year and 0 otherwise.	Compustat
ROA	Return on assets; the ratio of net income to average total assets.	Compustat
Cash Flow	Operating cash flows scaled by total assets.	Compustat
Foreign	Equals 1 if a firm reports any foreign income and 0 otherwise.	Compustat
Segments	The log of the total number of business segments.	Compustat
Report Lag	The log of the number of days between a client firm’s fiscal year-end and the audit report date.	Audit Analytics
Big 4	Equals 1 if the auditor firm is one of the Big 4 auditors and 0 otherwise.	Audit Analytics
Specialist	A measure of the auditor firm’s industry specialization (based on Dunn & Mayhew, 2004); equals 1 if the total sales audited by an auditor firm within an industry above 30% and 0 otherwise.	Compustat
Prior Relationship	Equals 1 if there is an existing working relationship between the auditor and client (i.e., if the auditor firm is not a new auditor of the client firm) and 0 otherwise.	Audit Analytics
Bargaining Power	Equals 1 if a client firm’s total assets are greater than the median of all the client firms of that auditor in the current year and 0 otherwise.	Compustat
Reporting Quality	A measure of reporting quality (based on Fung et al., 2017); equals 1 if (a) the firm-year net income before extraordinary items scaled by total assets lies in the interval [0, 0.005], (b) the firm-year change in net income before extraordinary items scaled by total assets lies in the interval [0, 0.005], or (c) the firm-year actual earnings per share less the analyst consensus forecast earnings per share lies in the interval [0, 1 cent] and 0 otherwise. We multiply the value by –1 so that higher values indicate higher reporting quality.	Compustat
Customer Secret	A measure of the trade secrets of the firm’s major customers, constructed following G. Chen et al. (2022) as the sales-weighted count of whether its major customers mention trade secret-related keywords (e.g., ‘trade secret’, ‘trade secrecy’) in 10-K filings.	SEC Edgar

) summarizes the definitions of all variables. We include year and firm fixed effects to account for time- and firm-invariant factors influencing audit fees and double-cluster standard errors by year and firm.

Table 1. Summary statistics.

Variable	N	Mean	SD	Median	1st Quartile	3rd Quartile
Audit Fee (million$)	73,517	1.356	2.479	0.500	0.153	1.449
Audit Fee (ln)	73,517	13.042	1.563	13.122	11.939	14.186
Redaction	73,517	0.079	0.270	0.000	0.000	0.000
Audit Tenure	73,517	1.678	0.644	1.609	1.099	2.197
Total Assets (billion$)	73,517	2.480	1.050	0.232	0.036	1.217
Size	73,517	5.191	2.735	5.447	3.594	7.104
Acc_Filer	73,517	0.593	0.518	1.000	0.000	1.000
Leverage	73,517	0.461	1.403	0.190	0.012	0.398
Liquidity	73,517	0.240	0.200	0.200	0.071	0.359
GC	73,517	0.393	0.488	0.000	0.000	1.000
Restatement	73,517	0.085	0.279	0.000	0.000	0.000
Loss	73,517	0.467	0.499	0.000	0.000	1.000
ROA	73,517	−0.133	0.334	0.010	−0.201	0.064
Cash Flow	73,517	−0.163	0.790	0.056	−0.076	0.118
Foreign	73,517	0.415	0.493	0.000	0.000	1.000
Segments	73,517	1.053	0.475	0.693	0.693	1.386
Report Lag	73,517	4.179	0.370	4.205	4.007	4.407
Big 4	73,517	0.510	0.500	1.000	0.000	1.000
Compensation	73,517	0.002	0.042	0.000	0.000	0.000
Financial	73,517	0.005	0.071	0.000	0.000	0.000
Business	73,517	0.020	0.142	0.000	0.000	0.000
Intangible	73,517	0.013	0.115	0.000	0.000	0.000
Other	73,517	0.013	0.113	0.000	0.000	0.000

All variable definitions are summarized in Table A1.

displays the summary statistics, where all continuous variables are winsorized at highest percentile levels. The mean (median) audit fee is 13.042 (13.122) with a standard deviation of 1.563, translating to approximately US$500,000 median in audit fees during the sample period. On average, around 8% of firm-year observations involve a redacted report. The remaining variables appear to follow normal distributions.

Table 2. Correlations.

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
1. Audit Fee		0.02 *	0.47 *	0.86 *	0.54 *	0.11 *	0.09 *	−0.06 *	0.03 *	−0.34 *	0.37 *	0.39 *	0.51 *	0.16 *	−0.39 *	0.45 *
2. Redaction	0.03 *		0.02 *	−0.02 *	−0.00	−0.03 *	−0.06 *	−0.01	−0.00	0.06 *	−0.06 *	−0.06 *	−0.01	0.04 *	0.02 *	0.03 *
3. Audit Tenure	0.45 *	0.01		0.38 *	0.23 *	0.02 *	0.02 *	−0.08 *	−0.04 *	−0.20 *	0.22 *	0.20 *	0.24 *	0.12 *	−0.15 *	0.22 *
4. Size	0.85 *	0.00	0.37 *		0.59 *	0.16 *	0.05 *	−0.11 *	−0.01	−0.46 *	0.50 *	0.50 *	0.43 *	0.18 *	−0.54 *	0.47 *
5. Acc_Filer	0.50 *	−0.00	0.22 *	0.53 *		−0.02 *	0.06 *	−0.09 *	−0.02 *	−0.33 *	0.35 *	0.36 *	0.31 *	0.10 *	−0.43 *	0.36 *
6. Leverage	−0.23 *	−0.02 *	−0.06 *	−0.37 *	−0.14 *		−0.01	0.16 *	0.05 *	0.06 *	−0.13 *	−0.07 *	−0.03 *	0.01	0.08 *	−0.02 *
7. Liquidity	0.01	−0.06 *	−0.01	−0.01	−0.01	−0.05 *		−0.08 *	−0.00	−0.22 *	0.25 *	0.18 *	0.16 *	−0.01	−0.05 *	−0.02 *
8. GC	−0.08 *	−0.01	−0.11 *	−0.17 *	−0.08 *	0.20 *	−0.07 *		0.05 *	0.16 *	−0.21 *	−0.17 *	−0.06 *	−0.07 *	0.16 *	−0.05 *
9. Restatement	0.02 *	−0.00	−0.03 *	−0.02 *	−0.02 *	0.02 *	−0.00	0.05 *		0.04 *	−0.05 *	−0.04 *	−0.00	−0.01	0.11 *	−0.03 *
10. Loss	−0.34 *	0.06 *	−0.21 *	−0.45 *	−0.31 *	0.17 *	−0.17 *	0.16 *	0.04 *		−0.86 *	−0.63 *	−0.20 *	−0.11 *	0.33 *	−0.18 *
11. ROA	0.45 *	−0.05 *	0.21 *	0.63 *	0.34 *	−0.34 *	0.20 *	−0.27 *	−0.03 *	−0.68 *		0.74 *	0.23 *	0.14 *	−0.36 *	0.21 *
12. Cash Flow	0.39 *	−0.01	0.14 *	0.59 *	0.25 *	−0.61 *	0.14 *	−0.25 *	−0.02 *	−0.35 *	0.64 *		0.22 *	0.15 *	−0.36 *	0.23 *
13. Foreign	0.50 *	−0.01	0.25 *	0.42 *	0.29 *	−0.13 *	0.09 *	−0.06 *	−0.00	−0.20 *	0.26 *	0.21 *		0.02 *	−0.23 *	0.22 *
14. Segments	0.20 *	0.02 *	0.14 *	0.23 *	0.10 *	−0.07 *	−0.01	−0.05 *	−0.00	−0.14 *	0.18 *	0.14 *	0.04 *		−0.10 *	0.08 *
15. Report Lag	−0.28 *	0.02 *	−0.09 *	−0.44 *	−0.34 *	0.20 *	0.01	0.16 *	0.12 *	0.27 *	−0.32 *	−0.27 *	−0.18 *	−0.08 *		−0.41 *
16. Big 4	0.45 *	0.03 *	0.22 *	0.47 *	0.35 *	−0.15 *	−0.07 *	−0.05 *	−0.03 *	−0.18 *	0.25 *	0.23 *	0.22 *	0.08 *	−0.36 *

Lower-triangular cells display Pearson’s correlations; upper-triangular cells show Spearman’s rank correlations. * indicates significance at the 5%. All variables are defined in Table A1.

shows Pearson correlations in the lower triangle and Spearman correlations in the upper triangle. Both sets of correlations reveal a positive and significant association between redaction and audit fees, thus supporting our primary prediction, though this remains indicative of the underlying relationship.

4. Results

4.1. Primary Results

Table 3. The effects of redaction on audit fees: Main tests.

Variable	Audit Fee	Audit Fee	Audit Fee	Audit Fee
	(1)	(2)	(3)	(4)
Redaction	0.126 ***	0.087 ***	0.044 ***	0.048 ***
	(0.021)	(0.017)	(0.009)	(0.009)
Audit Tenure	0.329 ***	0.158 ***	0.174 ***	0.165 ***
	(0.058)	(0.021)	(0.020)	(0.021)
Size	0.495 ***	0.508 ***	0.352 ***	0.347 ***
	(0.004)	(0.005)	(0.009)	(0.009)
Acc_Filer	0.144 ***	0.140 ***	0.239	0.221
	(0.030)	(0.025)	(0.191)	(0.191)
Leverage	0.024 ***	0.035 ***	0.027 ***	0.029 ***
	(0.005)	(0.004)	(0.004)	(0.004)
Liquidity	0.336 ***	0.426 ***	0.321 ***	0.341 ***
	(0.035)	(0.039)	(0.034)	(0.035)
GC	0.070	0.069 ***	0.038 ***
	(0.043)	(0.017)	(0.010)
Restatement	0.134 ***	0.101 ***	0.081 ***
	(0.020)	(0.013)	(0.011)
Loss	0.070 ***	0.072 ***	0.010	0.021 **
	(0.019)	(0.015)	(0.009)	(0.009)
ROA	–0.293 ***	–0.302 ***	–0.199 ***	–0.225 ***
	(0.031)	(0.029)	(0.019)	(0.020)
Cash Flow	−0.209 ***	−0.195 ***	−0.125 ***	−0.125 ***
	(0.009)	(0.008)	(0.007)	(0.007)
Foreign	0.452 ***	0.315 ***	0.149 ***	0.153 ***
	(0.016)	(0.013)	(0.016)	(0.017)
Segments	0.034 **	0.022 *	0.024 *	0.023 *
	(0.012)	(0.011)	(0.013)	(0.013)
Report Lag	0.452 ***	0.246 ***	0.241 ***
	(0.033)	(0.035)	(0.028)
Big 4	0.153 ***	0.237 ***	0.244 ***	0.244 ***
	(0.039)	(0.022)	(0.018)	(0.019)
Intercept	Yes	Yes	Yes	Yes
Fixed Effects	No	Industry, Year	Firm, Year	Firm, Year
Adjusted R2	0.809	0.840	0.913	0.911
N	73,517	73,517	73,517	73,517

All variable definitions are summarized in Table A1. Standard errors (in parenthesis) are double-clustered by year and firm. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

presents the main findings, where Audit Fee is regressed on the Redaction indicator along with the control variables. The first three columns display the results from regression (1): column (1) excludes fixed effects; column (2) contains industry and year fixed effects; and column (3), the most conservative model, includes fixed effects that control for time- and firm-invariant factors that might affect audit fees. Across all specifications, while the coefficient on Redaction decreases in magnitude from column (1) to column (3), it remains significant at conventional levels. In terms of economic significance, with a median audit fee of US$500,000 per firm-year in the sample period, the results in column (3) suggest that redaction is associated with an annual audit fee increase of about US$22,500 (5%). Given the average audit tenure of five years during the sample period, this results in overall savings of US$112,500. Finally, since GC, Restatement, and Report Lag are potential outcomes of the audit process, we exclude these variables in column (4) for a sensitivity analysis, which confirms the robustness of our results under this alternative specification.6

For control variables, we find results consistent with prior studies. For example, firms in which auditors have a longer tenure with their clients experience higher audit fees, consistent with the argument presented in Brandon and Mueller (2008) that longer-tenure auditors are subject to higher litigation risks. Firms with larger size (Size) and higher leverage (Leverage) and liquidity ratio (Liquidity) have higher audit fees, while those with higher ROA and Cash Flow incur significantly lower fees.7 The audit fees are also significantly higher for firms with more complex businesses, such as those with foreign income (Foreign) and more business segments (Segments), and among firms with longer reporting lag (Report Lag) and that are audited by one of the Big 4 auditors, who provide higher audit quality and/or face higher exposure to litigations (DeFond & Zhang, 2014). It is also worth noting that adding firm and year fixed effects results in a higher adjusted R²; however, some of the control variables lose their significance. Overall, results in Table 3 present evidence consistent with our first hypothesis that redacting firms incur significantly higher audit fees.

4.2. Additional Explanations

We present various sensitivity tests and explore different sources of redaction in

Table 4. The effects of redaction on audit fees: Sources of redaction.

Variable	Audit Fee	Audit Fee	Audit Fee
	(1)	(2)	(3)
# Redaction	0.020 ***
	(0.005)
Redaction All		0.025 ***
		(0.007)
Compensation			−0.023
			(0.039)
Financial			−0.033
			(0.028)
Business			0.048 **
			(0.018)
Intangible			0.038 **
			(0.017)
Other			0.012
			(0.017)
Controls	Yes	Yes	Yes
Intercept	Yes	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes	Yes
Adjusted R2	0.913	0.913	0.913
N	73,517	73,517	73,517

Redaction All is a measure based on 10-Qs, 10-Ks, and 8-Ks. The redacted information falls into three main categories: executive compensation (Compensation), contract-related redactions [financial services (Financial), business dealings with suppliers and customers (Business), and intangible assets such as license agreements, software, and patents (Intangible)], and unspecified items (Other). Standard errors (in parenthesis) are double-clustered by year and firm. ** and *** denote significance at the 5% and 1% levels, respectively. All other variable definitions are summarized in Table A1. Controls include the same set as in Table 3, model 3.

. First, we substitute the Redaction indicator with the number of redaction-related keywords found in a firm’s 10-Q or 10-K filings each year (# Redaction in column (1))8, and with a measure that includes 10-Qs, 10-Ks, and 8-Ks (Redaction All in column (2)). In both cases, the coefficients remain significant both statistically and economically. Second, we further differentiate between the types of redacted information and explore whether the nature of the redacted content has a differential impact on audit fees. In particular, we categorize the redacted information into three main groups: executive compensation (Compensation), contract-related subcategories [financial services (Financial), business dealings with suppliers and customers (Business), and intangible assets such as license agreements, software, and patents (Intangible)], and unspecified items (Other). This categorization follows G. Chen et al. (2022), who show that suppliers are more likely to redact required disclosures when their major customers possess proprietary information that needs protection (i.e., trade secrets, R&D intensity, and nondisclosure agreements). We classify the redaction information based on the information we collect from the 10-Q and 10-K filings.9

In column (3), we find that redactions involving business dealings with suppliers and customers (Business) and intangible assets such as license agreements, software, and patents (Intangible) exert the greatest influence on audit fees. Our results are intuitive and align with the varying audit implications of different types of redacted information. For example, redactions related to executive compensation are unlikely to affect audit fees because auditors can verify these amounts through internal payroll and human resources records, independent of public disclosure; thus, such omissions introduce little additional audit complexity or risk. Similarly, financial contracts often follow standardized terms and are already subject to extensive audit testing and third-party confirmations, so their redaction does not materially increase audit effort or perceived engagement risk. In contrast, redactions involving business dealings with suppliers and customers directly affect auditors’ understanding of a client’s operations and revenue sources. Concealing these contractual details eliminates the informational ‘spillover’ that typically reduces audit uncertainty, increasing both audit complexity and perceived risk. Redactions concerning intangible assets (i.e., licenses, software, or patents) pose even greater challenges, as valuation and ownership assessments rely on detailed disclosures. The lack of transparency in these areas heightens the risk of misstatement and litigation exposure, leading auditors to charge higher fees. Finally, redactions categorized as unspecified items likely consist of diverse or immaterial matters that do not systematically influence audit risk or cost. The findings in column (3) are also consistent with J. Chen et al. (2014), who show that auditors with multi-client supply chain experience charge lower fees because shared contract knowledge reduces information asymmetry.

4.3. Cross-Sectional Results

We perform different cross-sectional tests to provide further evidence of the mechanisms underlying the increased audit fees for redacting firms. First, to gauge if the increase is mainly attributable to heightened risks of material misstatement or an increase in the auditor’s reputational and litigation risks, we modify regression (1) by interacting Redaction with Big 4 (one of the Big 4 auditors), Audit Tenure (the length of the client–auditor relationship), Specialist (the auditing firm’s industry specialization), and Prior Relationship (the presence of a prior client–auditor relationship, which excludes new audit engagements), and tabulate the results in

Table 5. The effects of redaction on audit fees: Cross-sectional tests.

Variable	Audit Fee	Audit Fee	Audit Fee	Audit Fee	Audit Fee	Audit Fee
	(1)	(2)	(3)	(4)	(5)	(6)
Redaction	0.071 ***	0.082 ***	0.054 ***	0.145 ***	0.064 ***	0.088 ***
	(0.013)	(0.017)	(0.011)	(0.027)	(0.012)	(0.019)
Redaction * Big 4	–0.047 **
	(0.020)
Redaction * Audit Tenure		−0.073 ***
		(0.022)
Redaction * Specialist			−0.032 *
			(0.017)
Redaction * Prior Relationship				−0.119 ***
				(0.029)
Redaction * Bargaining Power					−0.040 **
					(0.016)
Redaction * Reporting Quality						−0.049 **
						(0.018)
Specialist			0.030 **
			(0.011)
Prior Relationship				0.001
				(0.023)
Bargaining Power					−0.088 ***
					(0.015)
Reporting Quality						−0.014
						(0.010)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Intercept	Yes	Yes	Yes	Yes	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.913	0.913	0.913	0.913	0.913	0.913
N	73,517	73,517	73,517	73,517	73,517	73,517
Wald Test (β1 + β2)	3.19 *	0.66	2.82	7.25 **	4.25 *	17.55 ***

***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. All variable definitions are summarized in Table A1. Controls include the same set as in Table 3, model 3.

, columns (1)–(4). All four interaction coefficients are significantly negative; this suggests that the increase in audit fees is significantly reduced among firms with specialist auditors, Big 4 auditors, and auditors with existing and longer relationships with the redacting firms. These findings indicate that the increased audit fees are likely to be driven by heightened reputational and litigation risks than the need for greater auditor effort due to higher risks of material misstatements. While Big 4 auditors could be more attuned to reputational and litigation concerns, their superior resources, expertise, and client-specific knowledge enable them to handle the perceived engagement risk from redacted information more effectively. As a result, their fee adjustments are smaller compared to those of non-Big 4 auditors (Brandon & Mueller, 2008).

Second, to determine whether higher audit fees are driven by auditors’ need to compensate for increased efforts or by the redacting firm’s demand for higher-quality assurance (DeFond & Zhang, 2014), we interact Redaction with the firm’s bargaining power and reporting quality, defined in Table A1. The results are summarized in Table 5, columns (5) and (6). The coefficients for both interaction terms are significantly negative, which indicates that firms with greater bargaining power and higher reporting quality experience a smaller increase in audit fees.10 These findings suggest that auditors increase their fees for redacting firms due to increased effort, but firms with more bargaining power and higher-quality reporting can mitigate these additional costs. These results are not necessarily inconsistent with those reported in columns (1)–(4); while auditors may be concerned about reputational and litigation risks for redacting firms, their status, specialization, and prior experience with the client allow them to manage engagement risk more efficiently, which in turn reduces the effort required.

Results in

Table 6. The effects of redaction on audit fees: Additional tests.

Variable	Audit Fee	Audit Fee	Audit Fee	Audit Fee	Audit Fee	Audit Fee	Audit Fee
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Redaction	0.034 ***	0.036 ***	0.036 ***	0.044 ***	0.044 ***	0.039 ***	0.036 ***
	(0.011)	(0.008)	(0.008)	(0.009)	(0.009)	(0.008)	(0.008)
Opacity	1.794 ***
	(0.537)
ICW Count		0.207 ***
		(0.015)
ICW			0.205 ***				0.205 ***
			(0.021)				(0.021)
Equity Offer				0.005			0.017 ***
				(0.007)			(0.004)
M&A					0.010		0.016 **
					(0.006)		(0.007)
Debt Issue						0.006	0.004
						(0.009)	(0.008)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Intercept	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.900	0.924	0.923	0.913	0.913	0.915	0.923
N	39,299	58,586	58,586	69,312	70,566	67,261	58,586

Opacity is the yearly average of daily bid-ask spread (based on Corwin & Schultz, 2012); ICW Count is the log of (1 plus the number of internal control weakness from the auditor); ICW equals 1 if there is at least one internal control weakness from the auditor opinion and 0 otherwise; Equity Offer equals 1 if firms have seasoned equity offerings and 0 otherwise; M&A equals 1 if there is at least one merger or acquisition in the fiscal year and 0 otherwise; and Debt Issue equals 1 if the firm has issued new bonds in the fiscal year and 0 otherwise. Standard errors (in parenthesis) are double-clustered by year and firm. ** and *** denote significance at the 5% and 1% levels, respectively. All other variable definitions are summarized in Table A1. Controls include the same set as in Table 3, model 3.

examine whether our main findings hold when including additional control variables commonly utilized in the audit fee literature (Danielsen et al., 2007; Kuang et al., 2021), where the auditor faces heightened engagement risk. In column (1), we add Opacity, measured as the yearly average of the daily bid-ask spread (following Corwin & Schultz, 2012), and find that redaction continues to have a significant impact on audit fees, even after accounting for the impact of an opaque information environment, as noted by Danielsen et al. (2007). Columns (2)–(6) introduce controls for the number of weaknesses (ICW Count) and the presence of internal control weakness (ICW), seasoned equity offerings (Equity Offer), mergers and acquisitions (M&A), and new debt issuance (Debt Issue). The final column includes all four indicators together and confirms that redaction remains a statistically and economically significant determinant of audit fees.

4.4. Sensitivity Tests

To alleviate concerns about potential endogeneity—particularly that firms audited by Big 4 auditors may be structurally different in size, complexity, or governance—we adopt several complementary approaches. Our baseline regressions include firm fixed effects, which absorb all time-invariant characteristics that could jointly influence both the decision to redact and audit pricing, ensuring that the estimated coefficient reflects within-firm variation over time. However, concerns may remain that our main finding is driven by correlated omitted variables that are associated with both the decision to redact and set audit fees. To mitigate this and other endogeneity concerns, we conduct the following robustness tests.

First, we re-estimate regression (1) using matched samples. We compare the characteristics of redacting and nonredacting firms and find that these two groups differ significantly along several dimensions, such as leverage, profitability, business complexity, and auditor characteristics. To ensure that our key result is not driven by systematic differences between the redacting and nonredacting firms, we construct matched samples using propensity scores or entropy balancing. To identify the propensity score-matched sample, we first regress the Redaction indicator on the control variables in regression (1) along with industry and year fixed effects and select the matched nonredacting firms using optimal matching. This method simultaneously selects all matches without replacement to minimize the total absolute difference in propensity scores across pairs, and has been shown to produce better-matched samples than alternative matching techniques (Rosenbaum, 2002; Stuart, 2010). When we compare the bivariate differences between nonredacting and redacting firms after the propensity score-matching, we find that the differences become statistically nonsignificant after matching (results not tabulated). We also construct a matched sample using entropy balancing based on the same set of control variables used in regression (1), where the differences between the redacting and nonredacting firms become statistically nonsignificant after matching.

Table 7. PSM and entropy balancing: Regressions using matched samples.

Variable	Audit Fee	Audit Fee
	(1) PSM	(2) Entropy Balancing
Redaction	0.027 *	0.038 ***
	(0.014)	(0.010)
Audit Tenure	0.164 ***	0.156 ***
	(0.022)	(0.019)
Size	0.335 ***	0.338 ***
	(0.014)	(0.009)
Acc_Filer	−0.115	0.133
	(0.124)	(0.125)
Leverage	0.029 **	0.029 ***
	(0.012)	(0.007)
Liquidity	0.392 ***	0.402 ***
	(0.084)	(0.053)
GC	0.039 ***	0.044 ***
	(0.011)	(0.009)
Restatement	0.097 ***	0.096 ***
	(0.019)	(0.016)
Loss	0.011	0.011
	(0.019)	(0.011)
ROA	−0.206 ***	−0.185 ***
	(0.038)	(0.023)
Cash Flow	−0.093 ***	−0.111 ***
	(0.020)	(0.011)
Foreign	0.133 ***	0.144 ***
	(0.036)	(0.022)
Segments	0.068 ***	0.047 **
	(0.021)	(0.017)
Report Lag	0.238 ***	0.241 ***
	(0.034)	(0.030)
Big 4	0.197 ***	0.234 ***
	(0.032)	(0.022)
Intercept	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes
Adjusted R2	0.914	0.917
N	9,304	73,517

Standard errors (in parenthesis) are double-clustered by year and firm. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. All variable definitions are summarized in Table A1.

presents the estimation results of regression (1) using the two matched samples. In both columns, the coefficient on Redaction remains significantly positive, which indicates that our key finding is probably not driven by differences in the observable characteristics between firms with and without redactions.

Second, to account for the potential effects of unobservable time-invariant firm characteristics on disclosure strategy and audit fees, we use a time-series design to examine the change in audit fees after the first-time firms file a redacted report. Specifically, for each firm with redactions, we define indicator variables: Current (equals 1 if the present year t is the year when the first redaction occurs and 0 otherwise), Before_2 [Before_1] (equals 1 if the present year t is two years [one year] before the first redaction and 0 otherwise), and After_1 [After_2+] (equals 1 if year t is one year [two or more years] after the first redaction and 0 otherwise). We then modify regression (1) by regressing Audit Fee on these indicators along with control variables and fixed effects.

The results are summarized in

Table 8. Time series analysis.

Variable	Audit Fee	Audit Fee	Audit Fee	Audit Fee
	(1)	(2)	(3)	(4)
Post Redaction	0.034 *		0.042 **
	(0.020)		(0.018)
Before_2		−0.001		0.001
		(0.020)		(0.018)
Before_1		0.017		0.023
		(0.020)		(0.017)
Current		0.069 ***		0.070 ***
		(0.022)		(0.016)
After_1		0.044 *		0.042 **
		(0.022)		(0.017)
After_2+		0.026		0.031 *
		(0.026)		(0.018)
Controls	Yes	Yes	Yes	Yes
Intercept	Yes	Yes	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes	Yes	Yes
Adjusted R2	0.913	0.913	0.893	0.893

Standard errors (in parenthesis) are double-clustered by year and firm. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Controls include the same set of variables as in Table 3, model 3. All variable definitions are summarized in Table A1.

. In the first two columns, the results are based on the full sample, which includes firms that have never redacted their filings. In column (1), Post Redaction equals 1 if the present year t is the year of the first redaction or after the year of the first redaction and 0 otherwise. Its coefficient is significantly positive, consistent with our main findings. Column (2) presents the findings from the time series model. The coefficients on Before_2 and Before_1 are both statistically nonsignificant, suggesting that there is no significant change in audit fees for firms in the first and second years prior to the first redaction. The coefficients on Current, After_1, and After_2+ are all significant and positive; the magnitude of the coefficients decreases over time, suggesting that firms experience a significant increase in audit fees in the year of the first redaction and that the significant increase persists over the next few years (though with decreasing magnitudes). More interestingly, inferences remain qualitatively similar and stronger if we repeat this analysis using a subsample that excludes firms that have never redacted filings in columns (3) and (4). Overall, our main finding does not seem to be explained by certain innate characteristics of the redacting firms.

Finally, to further mitigate correlated omitted variable concerns, we utilize the trade secrets of the firm’s major customers as an instrument for its own redaction decisions. G. Chen et al. (2022) show that customers with trade secrets tend to pressure suppliers to redact their filings to protect the customer’s proprietary information. Thus, customers’ trade secrets have a direct positive effect on the supplier’s redaction decisions but should not be associated with its audit fees. Similarly to G. Chen et al. (2022), we calculate a firm’s exposure to customer trade secrets (Customer Secret) as the sales-weighted count of whether its major customers mentioned trade secret-related keywords (e.g., “trade secret”, “trade secrecy”) in firms’ 10-K filings. We first estimate the first stage regression by regressing the Redaction indicator on Customer Secret and the control variables in regression (1). The estimation results in column (1) in

Table 9. Estimation using instrumental variable.

Variable	Redaction	Audit Fee
	(1)	(2)
Customer Secret	0.352 ***
	(0.044)
Redaction_Pred		0.138 **
		(0.066)
Controls	Yes	Yes
Intercept	Yes	Yes
Fixed Effects (Firm, Year)	Yes	Yes
Adjusted R2	0.409	0.228
N	73,517	73,517
Kleibergen–Paap RK LM statistic		7.768 **
Cragg–Donald Wald F statistic		25.368
		[16.38]

Standard errors (in parenthesis) are double-clustered by year and firm. ** and *** denote significance at 5%, and 1%, respectively. Controls include the same set of variables as in Table 3, model 3. All variable definitions are summarized in Table A1.

show that the coefficient on Customer Secret is significant and positive, consistent with the findings reported in G. Chen et al. (2022). We then replace Redaction in regression (1) with the predicted value of Redaction (Redaction_Pred) from the first stage model and find the coefficient on Redaction_Pred in column (2) to be significantly positive, thereby lending support to our hypothesis that redaction leads to an increase in audit fees.

5. Conclusions

This study examines how trade secrecy protection through redaction affects audit fees, offering new insights into the intersection of proprietary disclosure and audit effort. While auditors have access to confidential information, we document that firms that redact material information from public filings incur significantly higher audit fees, suggesting that redaction introduces complexity and perceived risk into the audit process. This relationship is further moderated by both auditor characteristics (e.g., specialization, tenure, and quality) and client factors (e.g., bargaining power and reporting quality), indicating that audit responses to redaction are context dependent. Theoretically, our findings extend the literature examining the economic consequences of trade secrecy protection by identifying auditors (beyond investors and regulators) as key stakeholders affected by redaction. We also extend the audit literature by finding that withholding information in mandatory filings can influence the auditor’s risk assessment and pricing decisions, even when the underlying information remains accessible. For practitioners, our results highlight that redacting proprietary information can increase audit costs, particularly when audit complexity or risk perception rises. For academics, we offer a framework linking proprietary disclosure choices to assurance costs, pointing to broader implications for how transparency, complexity, and risk are managed in financial reporting and audit engagements. Overall, our results offer valuable insights into how the redaction process influences auditors’ assessments and integration efforts as well as the pricing of different risks in audit fees.

While our study provides new evidence on how trade secrecy protection influences audit pricing, it poses several limitations that offer avenues for future research studies. First, we recognize that auditors’ responses to redaction may vary based on factors such as their individual risk tolerance, experience, or perception of the client’s credibility. We also do not directly observe auditors’ internal risk assessments or negotiation dynamics, which could further clarify how redaction affects audit effort and pricing decisions. Future studies could thus investigate these variations by examining office-level prior experience with clients that redact disclosures. Second, we acknowledge that our analysis is limited to U.S. public firms, and the results may not thus be generalizable to private firms, other countries, and different institutional settings with unique disclosure regulations or litigation risks. For example, in jurisdictions with stricter regulations or higher litigation risk, auditors might adopt a more cautious approach to assess their clients’ engagement risk, thus leading to higher audit fees for redacting firms. Future studies could explore cross-country comparisons or the effects of evolving disclosure rules (i.e., IFRS’s discussion on ‘Disclosure of Sensitive Information’) to deepen understanding of how confidentiality practices interact with audit quality, regulatory oversight, and financial transparency.