Auditor-Client Disagreements , Auditor Resignations , And Audit Fees Charged By Successor Auditors

This paper investigates the effects of auditor-client disagreement disclosure on auditor resignations and audit fee charged by successor auditors. Using a matched sample of auditor changes over the period 2003-2016, we find that auditor resignations are more often accompanied by auditor-client disagreements. We also find that Big 4 auditors are more likely to resign from their engagements when they disagree with their clients. Further, we document that successor auditors charge higher audit fees for firms that have disagreements with their predecessor auditors. Relative to non-Big 4 auditors, Big 4 successor auditors charge even higher audit fee for disagreement firms.


I. INTRODUCTION
n 1971, the Securities and Exchange Commission (SEC) required public companies to report any change of independent auditor along with a disclosure of any auditor-client disagreement on Form 8-K (SEC, 1971). 1 The 8-K regulatory requirements suggest that regulators believe that auditor-client disagreements are important "reportable events" that require public dissemination because market participants are likely to benefit from this information. Prior studies document a significantly negative market reaction surrounding the announcement date of auditor change or disagreement (Eichenseher, Hagigi & Shields, 1989;Smith & Nichols 1982;Dhaliwal, Schatzberg & Trombley, 1993;Beneish, Hopkins, Jansen & Martin, 2005) Despite the importance of auditor-client disagreements, relatively few studies focus on the antecedents and precedents of auditor-client disagreements. In this paper, we investigate how auditors react to auditor-client disagreements. Specifically, we examine whether auditorclient disagreements are more likely to associate with: (1) auditor resignations, (2) more frequent Big 4 auditor resignations, and (3) higher audit fees charged by successor auditors. Defond and Jiambalvo (1993) suggest that an auditor-client disagreement usually occurs when the auditor objects to a controversial accounting procedure that the management uses or proposes to use. On one hand, a disagreement might get resolved if a manager updates his position on an accounting issue to a compromised position during the auditorclient negotiation. On the other hand, a disagreement might get resolved because the auditor is willing to discuss and compromise on the interpretations of some imprecise accounting standards (Nelson et al. 2002). In other instances, even though auditor-client disagreement may not be resolved, auditors may not view such disagreements as adverse and therefore concede to client's reporting preferences. Thus, no switch of auditor occurs, and the auditor-client disagreement is not required to be disclosed in Form 8-K. However, when risk of material misstatement is high, auditor-client disagreement might precipitate auditor resignations.
Our results are based on a treatment sample of auditor-client disagreements and control sample of auditor changes over the period 2003-2016. We identify 1,761 disagreement sample observations (treatment firm observations) from Further, we examine the impact of disagreements on Big 4 auditors' client retention decisions by including a Big4_PRE indicator variable and an interaction variable between Big 4 and the disagreement indicator variable (BIG4_PRE and BIG4_PRE*DISAGREE) in our logistic regression model. Since large auditors have "deeper pockets" and higher reputation than small auditors, they are more likely to suffer greater losses from an audit failure (Feltham, Hughes & Simunic, 1991;Clarkson & Simunic, 1994). Thus, large auditors are more sensitive to engagement risk. Further, large auditors are less likely to acquiesce to client's preferred accounting choices. It is easier for large auditors to replace a risky client because they have a larger client base and face less intense competition in the audit industry (DeAngelo, 1981;Lennox, 1999;Ghosh & Lustgarten, 2006). Therefore, we predict and find that Big 4 auditors are more likely to resign from their engagements when they disagree with their clients.
Last, we investigate how successor auditors determine their audit fees when they accept new clients that have disagreements with their predecessor auditors. Prior studies document that audit fees are mainly determined by the cost of the external audit efforts and auditor's expected loss from litigation (Simunic, 1980). Relative to nondisagreement firms, disagreement firms are more likely to have earning manipulations and higher risk of material misstatement. Thus, successor auditors are more likely to charge disagreement firms with higher audit fee. We find that successor auditors are more likely to charge their clients with higher audit fees if these clients have auditor-client disagreements with their predecessor auditors. Relative to non-Big 4 auditors, Big 4 auditors charge even higher audit fees for new clients who have disagreements with their predecessor auditors.
We contribute to the auditor disagreement literature in several ways. First, we enrich the literature on auditor change and auditor-client disagreement by documenting auditor resignations are more often accompanied by auditor-client disagreement disclosure. Unlike other studies that examine auditor changes (Krishnan & Krishnan, 1997), we provide a theory explaining that predecessor auditors are more likely to resign from the engagements because the asymmetric information of disagreement may cause the successor auditor to underbid the engagements. Second, we extend the literature on Big 4 auditors' client retention decision in the context of auditor-client disagreement. We document that Big 4 auditors are more likely to resign from audit engagements when they disagree with their clients. This finding also contributes to the auditor change literature by confirming that Big 4 auditors have more stringent client retention policies than non-Big 4 auditors (Rama & Read, 2006). Finally, we document that successor auditors, including Big 4 and non-Big 4 auditors, are more likely to charge their new clients higher audit fee if their clients have auditor-client disagreement with predecessor auditors. This finding suggests that successor auditors perceive disagreement as a signal of high risk of material misstatement and litigation risk.
The rest of the paper is organized as follows. Section II provides a review of the auditor disagreement and auditor resignation literatures and develops the hypotheses. Section III describes the research design and the data. Section IV presents the testing results and Section V concludes the paper.

Auditor-Client Disagreement and Auditor Resignation
In 1971, the SEC required public companies to report any change of independent auditor and disclose any auditorclient disagreement that had occurred within 18 months prior to the auditor change on form 8-K (SEC, 1971). 2 The disagreement disclosure requirement was expanded in Accounting Series Release (ASR) 165 to include disagreement that occurred during the two most recent fiscal years and any subsequent interim period following the auditor change (Securities and Exchange Commission, 1974). In addition, in case of an auditor change, ASR 165 required the predecessor auditor to disclose any auditor-client disagreement in the auditor's report if disagreement is not resolved to the auditor's satisfaction and to submit a letter that states its agreement or disagreement regarding the content of the Form 8-K filed by the client (SEC, 1974). Antle and Nalebuff (1991) suggest that financial statements should be viewed as a joint statement from the auditor and management because it is the outcome of negotiations and bargaining between management and the auditor. During the auditor-client negotiation, a disagreement is more likely to be resolved if a manager updates his position on an accounting issue to a more compromised position. However, prior studies document that management might prefer adopting income-increasing accounting methods because of incentives to increase compensation and bonus awards, to increase job security, and to mitigate potential violation of debt covenants (Dhaliwal et al. 1993;Defond and Jiambalvo, 1994;Healy & Wahlen 1999). Even though management might prefer higher reported income, the firm bears substantial costs when challenging the auditor's position. For instance, because the negotiation process lengthens the auditing process, it inevitably leads to higher audit fee. Also, challenging auditor's position might result in an auditor resignation. Disclosing auditor resignation to the public might be costly because investors react negatively to auditor resignations (Wells & Loudder, 1997;Shu, 2000;Whisenant, Sankaraguruswamy & Raghunandan, 2003).
While the client prefers aggressive reporting, in sharp contrast, the auditor usually prefers conservative accounting choices as a protection against future litigation and the potential damages arising from aggressive reporting (DeFond & Subramanyam 1998;Reynolds & Francis, 2000). For instance, Francis and Krishnan (1999) document that auditors lower their threshold for issuing modified audit reports as a way to protect themselves from audit failure when their client has a high accrual. Also, auditors are more likely to resign from the engagement when litigation risk is high (Krishnan & Krishnan, 1997). However, auditors are also aware that excessive conservatism might lead to auditor dismissal. If management believes the incumbent auditor is more conservative than a successor auditor, management may want to dismiss the incumbent auditor with the expectation of replacing the incumbent auditor with a less conservative successor auditor.
An auditor-client disagreement usually occurs when auditor objects to a controversial accounting procedure that the management uses or proposes to use (Defond & Jiambalvo, 1993). The interpretation of complicated, imprecise, and controversial reporting standards might lead to ambiguity in reporting which could result in disagreement on the reporting of certain business transactions. Some of the auditor-client disagreements might get resolved through auditor-client negotiation because auditors are more likely to compromise in the auditor-client negotiations when accounting standards are imprecise (Nelson et al. 2002). In some other instances, even though auditor-client disagreement may not be resolved, auditors may not view such disagreements as adverse. However, when clients have high risk of material misstatement, auditors are less likely acquiesce to clients' positions and might even consider resigning from the engagements. Defond and Jiambalvo (1993) find that clients disagreeing with their auditors are more likely to manipulate earnings than other firms. Hirst (1994) documents that the risk of material misstatement increases when firms manage earnings. Thus, firms with auditor-client disagreements are more likely to have high risk of material misstatement. When a client's risk of material misstatement is high, the auditor usually expends more audit effort (e.g. extensive substantive test) to reduce audit risk to an acceptable level, which increases audit fees. Thus, fees are likely to be high for clients that disagree with their auditors. The Clute Institute On one hand, disagreements might provide auditors with incremental information which might lead to a reassessment of audit risk for the incumbent auditor. There is potential that the asymmetric information of disagreement between predecessor and successor auditors may cause the successor auditors to inappropriately underbid the audit. If the predecessor auditors anticipate the underbidding, they may resign rather than incur the costs of bidding. Also, if the risk of material misstatement is assessed too high for auditors after the disagreement, predecessor auditors may resign from the engagements because of the high litigation risk. On the other hand, disagreement might lead to auditor dismissal. Prior literature document that the clients may choose to dismiss the auditors if they receive qualified opinion (Krishnan, Krishnan, & Stephens, 1996). Thus, client disagreed with their auditors may dismiss their predecessor auditor to change a different auditor that will agree with them. Therefore, we posit the following hypothesis:

H1:
There is no significant relationship between auditor change and auditor-client disagreement.

Auditor-Client Disagreement and Large Auditors
Prior studies document that large auditors provide higher levels of audit quality than small auditors (e.g., DeAngelo 1981; Teoh & Wong, 1993;Krishnan, 2003;Wagner, 2012). Audit quality is higher for large auditors because they have reputational concerns and greater exposure to litigation risk than small auditors. Since large auditors have "deeper pockets" and more wealth at risk, they are likely to suffer greater losses from an audit failure (Feltham et al. 1991;Clarkson & Simunic, 1994). Therefore, large auditors are more sensitive to material misstatement in financial reports. In addition, because of deeper pockets, large auditors have more resources to invest in technology, personnel and training which results in higher audit quality. Thus, large auditors are more sensitive to engagement risk and are less likely to acquiesce to client's preferred accounting choices. DeAngelo (1981) and Lennox (1999) document that the audit market is segmented with a few large auditors having many clients while many small auditors compete for a smaller client base. Since large auditors have a large client base, it is rare that resigning from one risky client may have a large impact on their total revenue. However, a single client can constitute a major source of revenue for small auditors. In addition, Ghosh and Lustgarten (2006) find that large auditors face less intense competition in finding a new client because they dominate the oligopolistic segment of the audit industry. They also find that audit fee discounting is less extensive for large auditors in this segment. Relative to small auditors, large auditors are likely to find it easier to replace a risky client. Thus, large auditors are more likely to resign when they disagree with their clients. Therefore, we predict that Big 4 auditors have more stringent client retention policies than non-Big 4 auditors, which suggests that they are less likely to acquiesce to client pressure in case of a disagreement.
H2: Relative to non-Big 4 auditors, Big 4 auditors are more likely to resign from their audit engagements when they disagree with their clients on accounting matters.

Auditor-Client Disagreement and Audit Fees charged by Successor Auditors
In prior studies, the auditors' fee determination is explained in the following model: E(C)=cq +E(d|a, q)E(θ), where cq is the cost of the external audit efforts and E(d|a, q)E(θ) is the auditor's expected loss from litigation (Simunic, 1980). When a client's risk of material misstatement is high, the auditor usually expends more audit effort (e.g. extensive substantive test) to reduce audit risk to an acceptable level, which increases audit fees. In addition, higher risk of material misstatement and earning manipulations are more likely to increase the probability of litigation, which also increases audit fees. Relative to non-disagreement firms, disagreement firms are more likely to have higher risk, earning manipulations, and aggressive accounting reporting practice. Thus, successor auditors are more likely to charge disagreement firms with higher audit fee. The following hypothesis is tested:

Regression Models
To examine the association between auditor change and auditor-client disagreement, we perform a logistic regression analysis using a sample of disagreement firms and matched control firms. The dependent variable is auditor resignation that is an indicator variable equal to one if auditors resign from the engagements. Specifically, we estimate the following regression: = Natural Logarithm of total assets GCONCERN = 1 if the auditor opinion has a going-concern modification, and 0 otherwise; ACCRUALS = Firm's abnormal accruals in the fiscal year prior to the auditor change; REPEVENT = 1 if firms reported reportable events in their auditor change, and 0 otherwise; MODOP = 1 if the auditor opinion is modified opinion, and 0 otherwise; ICWEAKNESS = 1 if firms are identified with a material internal control weakness, and 0 otherwise; LOSS = 1 if firm's net earnings is less than zero, and 0 otherwise; REPLAG = Number of days from firms' fiscal year-end to audit report date. ALTZ = Decile 1 to 10 ranked by Altman Z score; 3 DISAGREE is the main variable of interest in this model. We control for audit opinion (MODOP) since Lys and Watts (1994) suggest that a modified opinion may increase the likelihood of auditor resignation if the qualification pertains to an error that occurred in prior year. We also control for reportable events (REPEVENT) and going concern opinions (GCONCERN) because Whisenant et al. (2003) find that these items discriminate between auditor resignations and dismissals. Following prior studies (Stice, 1991;Johnson, Khurana & Reynolds, 2002), we include variables TENURE and FEE to control for auditor independence and objectivity. We also control for audit report lag (REPLAG) because Mande and Son (2011) document that auditors are more likely to resign from clients with long audit report lag. In addition, motivated by prior studies (Stice, 1991;Krishnan & Krishnan 1997;Defond, Ettredge & Smith, 1997 Lastly, we examine how successor auditors react to auditor-client disagreement in term of audit pricing using a sample of treatment firms (firms have disagreement with their predecessor auditors) and control firms (firms have no disagreement with their predecessor auditors). We perform an ordinary least square regression analysis using the audit fee as the dependent variable. Specifically, we estimate the following regression: = / + 1 + 4 4_ * + 6 + 9 + > + @ + B + D + E 4_ + 1/ + 11 + 14 + In the above model, we investigate whether disagreement firms are more likely to pay higher audit fees to their successor auditors. We also include an interaction term BIG4_SUC*DISAGREE in the model to examine what is large auditors' pricing strategy on firms have disagreements with their predecessor auditors. We include measures to capture client firm size, characteristics and business risk (SIZE, LEVERAGE, CURASSETS, ROA, and LOSS), which prior studies have shown to be positively associated with audit fee (Simunic 1980;Newton & Ashton 1989). We also control for some auditor attributes including auditor resignation (RESIGN), auditor quality (BIG4_SUC), audit opinion (GCONCERN and MODOP), and engagement attribute (REPLAG).

Sample Selection
For our analyses, the initial sample includes all auditor changes observations from Audit Analytics' Auditor Changes file for the period 2003-2016. The going concern opinion data are from Audit Analytics' Audit Opinion file. Firms' financial data are from Compustat Annual files. Table 1 presents the sample selection for the disagreement sample and the control group sample for auditor change. First, we obtain 9,568 auditors change observations from Audit Analytics after it is merged with the Compustat data by CIK number. Second, we identify 1,761 disagreement observations (treatment firm observations) from the audit change database using variables of "Auditor_company_disagreement", "Dismissed_Auditor_Disagreement", and "Issue_Accounting". 4 Among the 1,761 disagreement observations, there are 578 auditor resignations and 1,183 auditor dismissals. For the 7,807 nondisagreement observations, there are 2,207 auditor resignations and 5,600 auditor dismissals. Third, we use the propensity score-matching method to match the disagreement observations with non-disagreement auditor change observations (control firm observations) without replacement by closest industry and year. The final sample includes 1,761 disagreement firm observations and 1,761 auditor change control firm observations. Notes: Table 1 presents the sample selection for the disagreement sample and the control group sample for auditor change. First, we obtain 9,568 auditor change observations from Audit Analytics after it is merged with the Compustat data by CIK from period 2003-2016. Second, we identify 1,761 disagreement observations (treatment firm observations) from the audit change database using variables of "Auditor_company_disagreement", "Dismissed_Auditor_Disagreement", and "Issue_Accounting". Third, we use the propensity score-matching method to match the disagreement observations with non-disagreement auditor change observations (control firm observations) without replacement by closest industry and year. The final sample includes 1,761 disagreement firm observations and 1,761 auditor change control firm observations. Table 2 presents descriptive statistics of all variables for the disagreement sample and the control auditor change sample. The result shows that disagreement firms are more likely to have audit resignation than control firms. We find that disagreement firms have significantly longer audit report lag and pay higher audit fee. Fee might be higher because of greater audit effort to cover added litigation risk. 5 Auditors usually input more audit effort (e.g. extensive substantive test) and need more time to perform the engagement when risk of material misstatement is high. Disagreement firms are more likely to receive modified audit opinion and going concern opinion, which suggests high risk of material misstatement because their financial statements are more likely to contain material misstatements. Iyer and Rama (2004) find that management is more likely to employ income-increasing accounting practices when audit tenure is short. Prior studies document that poor performance increases the likelihood that managers make aggressive accounting and reporting choices (Petroni, 1992;Sweeney, 1994;Beneish, 1997). Since both income-increasing accounting practices and aggressive accounting choices might lead to higher risk of material misstatements, disagreement firms are associated with higher risk because they have shorter audit tenure and higher probability of loss. In addition, disagreement firms have a significantly higher probability of reportable events and internal control weakness, suggesting that these firms have less effective and less reliable internal control system. In a word, the results suggest that disagreement firms have significantly higher risk of material misstatement than control firms in the year prior to the auditor change.

Copyright by author(s); CC-BY 22
The Clute Institute Notes: All firm financial variables are defined and calculated in the fiscal year prior to the auditor change. They are winsorized at the 1% level each tail to deal with outlier observations. Please see Appendix A for the variable definitions. *,**, and *** denote significance at 10%, 5% and 1% levels, respectively.
In Table 3, we present the Pearson correlations between the variables used in the regression. We document positive and significant correlations between auditor resignation and auditor-client disagreement.  Table 4 reports the results from the logistic model that estimates the effect of auditor-client disagreements on auditor resignations. The dependent variable is coded 1 for auditor resignations and 0 for auditor dismissal. We report two regression models for the resignation analysis in Table 4. In both models, we find a significantly positive relation between auditor-client disagreement and auditor resignation. The results provide evidence that auditors are more likely to resign from their engagements when they have disagreements with their clients. In model (2), the coefficients for FEE, GROWTH, and TENURE are significantly negative, suggesting that auditors are less likely to resign from clients who pay higher audit fee, clients who are growing, and clients with longer tenure. The significant negative coefficients on internal control weakness suggest that client firms are more likely to dismiss their auditors in cases of receiving an internal control weakness report. The probability that an auditor will resign rather than be dismissed is higher following the issuance of a going-concern opinion. Consistent with Mande and Son (2011), we find that auditors are more likely to resign from clients with long audit report lag. Notes: This table shows the results of logistic regression of auditor resignation and auditor-client disagreement. All control variables are measured in the fiscal year prior to the auditor change. Financial variables are winsorized at the 1% level each tail to deal with outlier observations. Please see Appendix A for the variable definitions. *,**, and *** denote significance at 10%, 5% and 1% levels, respectively. Table 5 presents the result of the logistic regression of Big 4 auditor resignation and auditor-client disagreement. By including the interaction variable (BIG4_PRE*DISAGREE) in the regression, we test whether Big 4 auditors are more likely to resign from their audit engagements when they disagree with their clients on accounting matters. The coefficient of the interaction term BIG4_PRE*DISAGREE is significantly positive while the coefficient on variable Big4_PRE is negatively significant. The results suggest that Big 4 auditors are less likely to resign from their client than non-Big 4 auditors. However, in circumstances of disagreement, Big 4 auditors are more likely to resign from their audit engagements than non-Big 4 auditors. One explanation for this finding is that Big 4 auditors become more conservative after Sarbanes-Oxley Act and more selective in accepting new client (e.g. larger clients, low risk clients), which results in less auditor resignations (Rama & Reed, 2006). However, in case of audit-client disagreement, Big 4

Auditor-Client Disagreement and Audit Fees
In Table 6, we investigate the effect of auditor-client disagreement on audit fees charged by successor auditors between disagreement firms and non-disagreement firms. In Model (1), the coefficient of DISAGREE is positively significant after controlling for firm level characteristics, business risk factors, and audit and engagement attributes. We find that successor auditors are more likely to charge their clients with higher audit fees if these clients have auditor-client disagreements with their prior auditors. The results suggest that disagreement firms are more likely to have higher risk of material misstatement and litigation risk than control firms, thus successor auditors incorporate this risk factor in their pricing and charge higher audit fees. In Model (2), we include an interaction term of BIG4_SUC*DISAGREE to investigate how Big 4 auditors will determine their audit fees for clients with audit-client disagreements. We have a positively significant coefficient for the interaction term, which shows that Big 4 successor auditors are more likely to charge even higher audit fees for firms with prior auditor-client disagreements. Prior studies document that Big 4 auditors face less intense competition in finding a new client and audit fee discounting is less extensive for large auditors (Ghosh & Lustgarten, 2006). Thus, Big 4 auditors are more likely to charge higher audit fees for disagreement firms to compensate their potential high litigation risk. Consistent with prior studies, we find auditors are more likely to charge higher audit fees for larger clients, clients with high business risk and audit risk. The audit fees charged by Big 4 auditors are higher than non-Big 4 auditors. Notes: This table shows the results of regression of audit fees charged by successor auditors and auditor-client disagreement. All control variables are measured in the fiscal year after the auditor change. Financial variables are winsorized at the 1% level each tail to deal with outlier observations. Please see Appendix A for the variable definitions. *,**, and *** denote significance at 10%, 5% and 1% levels, respectively.

V. CONCLUSION
This paper investigates the effects of auditor-client disagreement disclosure on auditor resignations and audit fee charged by successor auditors. This paper studies a sample of disagreement firms and a control sample of auditor change firms from the period 2003 to 2016. We document that firms with auditor-client disagreement have high risk of material misstatement. Controlling for other known factors affecting the likelihood of auditor resignation, we find that auditor resignations are more often accompanied by auditor-client disagreement disclosure. More importantly, we also find that disagreement has a greater negative impact on Big 4 auditors' client retention strategy. Relative to non-Big 4 auditors, Big 4 auditors are more likely to resign from their audit engagements when they have a disagreement with their clients. Lastly, we document that disagreements also have some impacts on audit pricing for successor auditors. We find that successor auditors are more likely to charge higher audit fees for disagreement firms and the audit fees charged by Big 4 successor auditors are even higher. The Clute Institute

Variables
Definitions RESIGN equals to 1 if auditors resigned from the engagements, and 0 otherwise. DISAGREE equals to 1 if firms are identified with disagreement in the auditor change database, and 0 otherwise. BIG4_PRE equals to 1 if firms' predecessor auditor is a Big 4 auditor, and 0 otherwise. BIG4_SUC equals to 1 if firms' successor auditor is a Big 4 auditor, and 0 otherwise. FEE Audit service fees as a percentage of total assets. LEVERAGE The ratio of total debt to total asset. GROWTH Growth in sales (scaled by prior year sales) in the year prior to the auditor change. TENURE Number of years of auditor tenure. SIZE Natural Logarithm of total assets. GCONCERN equals to 1 if the auditor opinion is a going-concern modification, and 0 otherwise. ACCRUALS Firm's abnormal accruals in the fiscal year prior to the auditor change. REPEVENT equals to 1 if firms reported reportable events in their auditor change, and 0 otherwise. ICWEAKNESS equals to 1 if firms are identified with a material internal control weakness, and 0 otherwise. MODOP equals to 1 if the auditor opinion is modified opinion, and 0 otherwise. LOSS 1 if firm's net earnings is less than zero, and 0 otherwise. REPLAG Number of days from fiscal year-end to audit report date. ALTZ Decile 1 to 10 ranked by Altman Z score. ROA Income before extraordinary items divided by total assets.