The Effects of Accounts with High Audit Risk on Auditor–Client Disagreement: Evidence from Korea

Abstract

This study examines the association between account-level inherent risk and auditor–client disagreement. To measure disagreement directly, we introduce a novel proxy: the absolute magnitude of the gap between pre-audit and post-audit net income, constructed from unique disclosure data available through the Korean KIND system. We specifically focus on accounts characterized by elevated inherent risk, including accounts receivable, inventory, investments in subsidiaries, defined benefit obligations, and derivatives. Our empirical results reveal heterogeneous associations that reflect competing theoretical tensions. The relative magnitudes of traditional operational accounts–specifically accounts receivable and inventory–as well as defined benefit obligations are significantly and negatively associated with auditor–client disagreement. In contrast, the magnitude of complex valuation accounts, particularly investments in subsidiaries, is positively associated with disagreement. We interpret these divergent findings as follows. The negative associations likely reflect the constraining effect of modern audit technologies on traditional high-risk accounts, which standardize audit procedures and thereby limit managerial discretion. The positive association, conversely, underscores the inherently subjective nature of complex fair-value estimates, which remain susceptible to auditor–client friction. Taken together, this study shifts the analytical focus from firm-level determinants to account-level risk, demonstrating that the underlying economic nature of an account systematically shapes the extent of audit negotiations.

1. Introduction

This study examines the relationship between account-level audit risk and auditor–client disagreement. Specifically, we analyze whether the relative magnitude of five accounts identified in prior research and practice as carrying high inherent risk—accounts receivable, inventory, investments in subsidiaries and associates, net defined benefit obligations, and derivative financial instruments—is systematically associated with the likelihood and magnitude of auditor–client disagreement. Disagreement is measured as the difference between pre-audit net income disclosed through mandatory timely filings, and post-audit net income reported in the audited financial statements. Accordingly, this study addresses the following research question: To what extent does the relative balance-sheet magnitude of specific high-inherent-risk accounts influence the incidence and magnitude of auditor–client disagreement arising during the pre-issuance financial reporting process?

Under the audit risk model (ISA 200), an auditor establishes a target level of acceptable audit risk at the outset of engagement planning, taking into account the client’s business environment and the information needs of stakeholders (IAASB, 2020). The auditor evaluates inherent risk (ISA 315) and, if planning to rely on the operating effectiveness of internal controls, assesses control risk at both the financial statement level and the individual account balance level, and calibrates the target level of detection risk accordingly (IAASB, 2019). When inherent risk is elevated for a specific account, the auditor reduces acceptable detection risk by intensifying substantive procedures and expending greater audit effort. Accounts characterized by high estimation uncertainty and valuation complexity are particularly susceptible to material misstatement arising from management’s subjective judgment. Consistent with this reasoning, Simunic (1980) provides foundational evidence that audit fees—a widely accepted proxy for audit effort—increase with client risk, reflecting the greater volume of work required to achieve acceptable assurance. When management’s incentives to report favorable outcomes conflict with an auditor’s obligation to apply professional skepticism to subjectively valued accounts, disagreements over the appropriate accounting treatment become more likely. In addition, as the risk of material misstatement associated with a specific account increases, auditors face heightened litigation exposure, and prior research shows that auditors respond to this litigation risk by improving audit quality, strengthening audit planning, and charging higher fees (Krishnan & Krishnan, 1997).

Despite its conceptual importance, auditor–client disagreement remains an understudied phenomenon, owing primarily to the difficulty of direct measurement. Traditional approaches rely on discretionary accruals as an indirect proxy for audit quality; however, because accrual estimates are derived from audited financial statements, they conflate management’s initial reporting choices with the outcome of auditor–client negotiations and are subject to well-documented measurement error. To address this limitation, Darvishi et al. (2026) adopt financial reporting quality as a proxy for auditor–client disagreement and construct a direct measure using the difference between pre-audit net income and post-audit net income—a methodological approach that more cleanly isolates the adjustment attributable to the audit process.

The Korean capital market offers an institutional setting uniquely suited to implementing this measurement approach. Listed firms on the KOSPI and KOSDAQ markets internally finalize their financial results prior to the completion of the external audit. Under the Korea Exchange’s Electronic Disclosure System (KIND), firms are required to publicly disclose any material change in their earnings structure—defined as a variation of 30% or more in sales revenue, operating income, or net income relative to the prior fiscal year (15% for large corporations)—under the mandatory filing category “Sales Revenue or Profit Structure Variation.” Specifically, the disclosure mandates that firms report not only the percentage variances but also the preliminary, un-audited absolute monetary figures for sales, operating income, and net income. These self-reported provisional figures serve as the precise pre-audit baseline in this study. This regulatory requirement generates a rich source of pre-audit earnings disclosures that, when matched with post-audit figures from audited annual reports, permits the construction of a reliable, account-level measure of auditor–client disagreement. Specifically, the disclosure mandates that firms report not only the percentage variances but also the preliminary, un-audited absolute monetary figures for sales, operating income, and net income. These self-reported provisional figures serve as the precise pre-audit baseline in this study. This study manually collects these pre-audit disclosures from KIND and calculates the difference between provisional pre-audit net income and audited post-audit net income as the primary measure of auditor–client disagreement.

While prior studies have advanced our understanding of firm-level determinants of auditor–client disagreement—such as board independence, internal control over financial reporting (ICFR) quality, and debt covenant pressure—the account-level risk characteristics that are associated with such disagreements remain largely unexplored (Simunic, 1980; Simon & Francis, 1988). This gap is consequential: auditors plan and execute procedures at the individual account balance and assertion levels, and inherent risk is rarely distributed uniformly across the financial statements. Grounded in the audit risk model and agency theory, this study argues that accounts with greater inherent risk, valuation complexity, and management estimation discretion are more likely to become the locus of negotiation between auditors and clients. Accordingly, we focus on five accounts with well-documented high audit risk: accounts receivable, inventory, investments in subsidiaries and associates, net defined benefit obligations, and derivative financial instruments. Throughout this study, the term “account” is used interchangeably with “financial statement line item” to refer to specific items presented in the financial statements. The selection of these five specific accounts is theoretically and practically grounded in regulatory enforcement actions and audit practice. According to common inspection findings by the PCAOB and frequently reported Critical Audit Matters (CAMs), audit deficiencies and heightened risks are consistently concentrated in revenue recognition (intimately linked to accounts receivable), inventory, and complex accounting estimates such as fair value measurements for investments and derivatives (PCAOB, 2023). By selecting a mix of traditional high-volume accounts and highly complex, estimate-driven accounts, this study captures a comprehensive spectrum of account-level inherent risk.

Our empirical analysis yields several noteworthy results. First, traditional high-risk items, such as accounts receivable, inventory, and net defined benefit obligations, exhibit a significant negative association with auditor–client disagreement. In contrast, investments in subsidiaries and associates—an account characterized by high valuation complexity—show a significant association with disagreement. Collectively, these findings demonstrate that auditors respond heterogeneously to the inherent risk of specific accounts, and that this differential response is directly observable in pre-audit to post-audit earnings adjustments.

This study makes three contributions to the extant literature. First, it extends research on auditor–client disagreement from the firm level to the account level, providing direct evidence on how account-specific inherent risk and complexity translate into observable negotiation outcomes. Second, it addresses a significant methodological limitation of prior studies by employing a direct, disclosure-based measure of auditor–client disagreement—the pre-audit to post-audit net income gap—that overcomes the identification problems inherent in accrual-based proxies. Third, it documents a theoretically meaningful asymmetry: traditional high-risk accounts with mature audit responses (accounts receivable, inventory, and net defined benefit obligations) are associated with lower disagreement, whereas a newly prominent, complex account (investments in subsidiaries and associates) is associated with higher disagreement—a distinction with practical implications for audit planning and standard-setting.

The remainder of this paper is organized as follows. Section 2 reviews the relevant literature and develops the hypotheses. Section 3 describes the research design and variable measurement. Section 4 reports the empirical results, including robustness tests. Section 5 discusses the findings in light of the theoretical framework and prior literature. Section 6 concludes.

2. Literature Review and Hypotheses Development

2.1. Motivation and Background

This study’s conceptual foundation integrates the audit risk model and agency theory. Under the audit risk model, audit risk is defined as the joint product of inherent risk, control risk, and detection risk (Simunic, 1980). To achieve an acceptably low level of overall audit risk, auditors respond to elevated assessments of inherent and control risks by reducing the tolerable level of detection risk—operationalized through more extensive and rigorous substantive procedures (Houston et al., 1999). Crucially, this risk-calibration process operates most acutely at the individual account balance and assertion levels. Accounts characterized by high estimation uncertainty and valuation subjectivity—such as investments in subsidiaries, derivatives, and defined benefit obligations—structurally amplify assertion-level inherent risk, demanding greater audit judgment and intensifying the potential for reporting conflicts over financial statement assertions (Christensen et al., 2012; DeFond & Zhang, 2014).

Agency theory provides a complementary perspective to explain why such conflicts materialize into observable disagreements. Management, acting as agents for shareholders, faces inherent incentives to exploit accounting discretion to maximize compensation or obscure poor economic performance (Jensen & Meckling, 1976). Such opportunistic behavior is most feasible in accounts where accounting standards afford significant measurement latitude (DeFond & Zhang, 2014). Conversely, external auditors bear a professional and legal obligation to challenge aggressive accounting estimates that deviate from underlying economic reality. When management’s optimistic reporting incentives collide with the auditor’s required professional skepticism over subjectively valued accounts, a structural disagreement arises (DeFond & Jiambalvo, 1993). This conflict is ultimately resolved through financial statement adjustments, which are directly observable in the Korean capital market as the measurable gap between pre-audit timely disclosures filed through KIND and final post-audit net income reported in audited financial statements.

Against this theoretical backdrop, this study examines five accounts with well-documented audit risk characteristics—accounts receivable, inventory, investments in subsidiaries and associates, net defined benefit obligations, and derivative financial instruments—as the primary vehicles through which account-level inherent risk translates into auditor–client disagreement. The theoretical logic developed in this section motivates the specific hypotheses formalized in Section 2.3.

2.2. Literature Review

This study draws on the audit risk model and agency theory to motivate its hypotheses. Under the traditional audit risk model (Simunic, 1980), auditors respond to elevated inherent risk at the account level by intensifying substantive verification procedures. Accounts characterized by high estimation uncertainty, fair-value subjectivity, or contractual complexity—such as investments in subsidiaries and associates, derivatives, and net defined benefit obligations—amplify inherent risk at the assertion level, thereby increasing the scope and depth of required audit work. From an agency theory perspective, managers facing earnings-based incentives may exploit the estimation discretion embedded in such accounts to achieve their reporting objectives. Auditors, serving as external monitors, are obligated to challenge accounting positions that deviate from applicable standards. When management’s preferred estimates conflict with auditors’ more conservative assessments, disagreement ensues (DeFond & Jiambalvo, 1993).

Crucially, the relationship between inherent risk and auditor–client disagreement is subject to competing theoretical tensions. Although agency theory predicts a positive association—driven by managerial opportunism and estimation uncertainty—the progressive standardization of verification procedures for traditional operational accounts through modern audit technologies may constrain managerial discretion, potentially giving rise to a negative association. Accordingly, the directional relationship between specific high-risk accounts and auditor–client disagreement ultimately remains an empirical question.

A substantial body of empirical research establishes that account-level risk is consistently associated with audit effort and fees. Simunic (1980) provides foundational evidence that auditors increase fees—reflecting greater audit effort—as client risk rises. Subsequent studies confirm that accounts receivable and inventory, owing to their valuation complexity and susceptibility to earnings management, are persistently associated with greater audit effort and higher fees (Simon & Francis, 1988; Widmann et al., 2021; Listalia & Suryaningrum, 2023). Extending this line of inquiry, Bowlin (2011) demonstrates experimentally that auditors allocate significantly greater resources to high-inherent-risk accounts relative to their low-risk counterparts, while Detzen et al. (2024) find that account-level risk materially shapes auditors’ within-engagement resource allocation decisions. With respect to derivatives, Ranasinghe et al. (2023) report that, although effective hedging reduces overall client business risk, auditors impose an additional risk premium when derivative complexity is high—underscoring that instrument-level complexity independently affects audit effort. Collectively, these findings affirm that auditors respond heterogeneously to account-level risk, providing the evidentiary foundation for examining whether such risk is also systematically associated with auditor–client disagreement.

Despite its importance, the auditor–client disagreement literature remains relatively sparse, largely owing to measurement challenges. Early studies relied on discretionary accruals as an indirect proxy; however, because accrual estimates are derived from audited financial statements, they cannot isolate management’s initial reporting assertions from the outcome of auditor–client negotiations (Sohn et al., 2010). To address this limitation, Sohn et al. (2010) introduce a direct measure—the difference between pre-audit net income, disclosed through timely regulatory filings, and post-audit net income, reported in audited financial statements—and demonstrate that greater board independence is associated with lower disagreement. DeFond and Jiambalvo (1993) show that debt covenant pressure and financial leverage are positively associated with auditor–client disagreement, consistent with the prediction that managers under financial distress resist downward audit adjustments. More recently, Kim (2024) finds that the functional expertise of internal control over financial reporting (ICFR) managers is negatively associated with auditor–client disagreement, as enhanced reporting quality narrows the gap between pre-audit and post-audit figures.

Taken together, prior research has predominantly examined firm-level determinants of auditor–client disagreement—including board structure, ICFR quality, and financial leverage—while largely neglecting account-level risk characteristics. This gap is notable given that auditors plan and execute procedures at the account balance and assertion levels, and inherent risk rarely distributes uniformly across the financial statements. To address this limitation, the present study examines whether the relative magnitude of five specific high-inherent-risk accounts—accounts receivable, inventory, investments in subsidiaries and associates, net defined benefit obligations, and derivatives—is systematically associated with both the likelihood and magnitude of auditor–client disagreement, as formalized in Section 2.3.

2.3. Hypotheses Development

The auditor’s assessment of the risks of material misstatement at the assertion level serves as the basis for determining the nature, timing, and extent of further audit procedures. Prior research suggests that specific balance sheet accounts carry inherent risks due to their complexity, subjectivity in valuation, and susceptibility to manipulation (Simunic, 1980; Simon & Francis, 1988). When auditors encounter accounts with high inherent risk, they exercise heightened professional skepticism and demand more rigorous audit evidence. This process often leads to a divergence between the auditor’s conservative view and management’s optimistic estimates, resulting in auditor–client disagreement.

Crucially, the association between account-level inherent risk and auditor-client disagreement is subject to competing theoretical tensions. From an agency theory perspective, the magnitude of high-risk accounts may exacerbate disagreement due to heightened estimation uncertainty and managerial opportunism. Conversely, from an audit maturity perspective, the verification of traditional operational accounts (e.g., accounts receivable and inventory) has become highly standardized through modern audit technologies, which may severely constrain managerial discretion and structurally reduce the scope for disagreement. Given these opposing forces, the directional association between specific high-risk accounts and auditor-client disagreement ultimately remains an empirical question.

Accounts receivable represents a significant portion of a firm’s assets and is directly linked to revenue recognition, a primary area for audit risk. Traditionally, verifying the existence of receivables was a labor-intensive process. While advancements in audit technology—such as automated confirmation platforms and data analytics (Huang et al., 2022)—have streamlined the verification of the existence assertion, the valuation assertion remains a source of significant estimation uncertainty and potential disagreement.

The valuation of accounts receivable depends heavily on the estimation of the allowance for doubtful accounts. This estimate requires management to make subjective judgments regarding the creditworthiness of customers, future economic conditions, and historical collection rates (Kulikova et al., 2015). Managers, incentivized to report higher earnings, may be inclined to underestimate bad debt expenses. Conversely, auditors, adhering to the principle of conservatism, are likely to challenge these optimistic estimates, especially in periods of economic uncertainty. This discrepancy in judgment regarding the realizability of receivables is likely to manifest as a measurable disagreement between the auditor and the client. Therefore, we posit that as the magnitude of accounts receivable increases, the incidence and magnitude of such disagreements will rise.

Under the audit risk model, the persistent estimation uncertainty embedded in the allowance for doubtful accounts elevates assertion-level inherent risk, compelling auditors to lower acceptable detection risk and intensify confirmatory and analytical procedures (Simunic, 1980). From an agency theory perspective, management facing earnings-based incentives has strong motives to underestimate bad debt provisions, creating a structural conflict with auditors’ obligation to apply professional skepticism to subjectively valued receivables (Jensen & Meckling, 1976; DeFond & Jiambalvo, 1993). Accordingly, we hypothesize:

H1. 

Ceteris paribus, the level of accounts receivable is systematically associated with auditor-client disagreement.

Inventory represents a second category of accounts characterized by substantial inherent risk, owing to the complexity of valuation and its direct link to on the cost of goods sold. While modern audit technologies have streamlined the verification of physical existence, the valuation assertion—specifically, the determination of Net Realizable Value (NRV) and the identification of obsolete or slow-moving items—remains highly subjective. Under agency theory, managers face strong incentives to delay recognizing inventory write-downs in order to maximize short-term reported earnings, thereby securing compensation bonuses and avoiding debt covenant violations (DeFond & Jiambalvo, 1994). Nelson et al. (2002) provide direct empirical evidence, drawn from auditors’ own accounts, that inventory valuation is among the most frequently contested areas of earnings management, as it affords managers considerable discretionary latitude. Conversely, auditors bear significant litigation and reputational exposure if overstated inventory balances are subsequently written off. Driven by professional skepticism and the conservatism principle, auditors routinely challenge optimistic managerial estimates of future sales prices and inventory recoverability. Under the audit risk model, this combination of high estimation uncertainty and strong managerial reporting incentives elevates assertion-level inherent risk, lowering the acceptable level of detection risk and intensifying substantive audit procedures (Simunic, 1980; Houston et al., 1999). This structural conflict between the agent’s reporting incentives and the auditor’s verification obligations creates fertile conditions for contested adjustments. Accordingly, we hypothesize:

H2. 

Ceteris paribus, the level of inventory is systematically associated with auditor–client disagreement.

Investments in subsidiaries and associates are inherently prone to auditor–client disagreement due to the substantial subjectivity involved in impairment testing. When objective evidence of impairment exists under K-IFRS, management must estimate the recoverable amount using complex discounted cash flow (DCF) models. Agency theory suggests that managers exploit the unverifiable nature of Level 3 fair value inputs—such as long-term growth rate projections and discount rate assumptions—to avoid recognizing impairment losses that would materially depress reported earnings and signal poor strategic performance to shareholders (Ramanna & Watts, 2012). Ramanna and Watts (2012) demonstrate empirically that managers with greater reporting flexibility under unverifiable fair value standards are significantly more likely to avoid timely impairment recognition. Griffith et al. (2015) and Barr-Pulliam et al. (2026) further show that auditors encounter considerable difficulty when verifying such complex, forward-looking estimates, frequently requiring the involvement of external valuation specialists and intensifying audit scrutiny. Because management’s optimistic valuation assumptions directly conflict with the auditor’s obligation to apply conservative and independent judgment, higher balances in these investment accounts amplify negotiation friction. Under the audit risk model, the unverifiable nature of Level 3 inputs and the consequent inability to reduce inherent risk through observable market benchmarks lower the acceptable level of detection risk, compelling auditors to deploy intensive valuation procedures and specialist involvement (Simunic, 1980; Griffith et al., 2015). Accordingly, we hypothesize:

H3. 

Ceteris paribus, the level of investments in subsidiaries and associates is positively associated with auditor–client disagreement.

Accounting for defined benefit (DB) plans introduces substantial estimation uncertainty due to its heavy reliance on actuarial assumptions, including discount rates, projected salary increases, and employee turnover rates (IAS 19, Employee Benefits). Because even marginal changes in these long-term assumptions can induce material shifts in reported financial position, inherent risk is particularly elevated for this account. Auditors are professionally obligated to rigorously evaluate whether these assumptions are anchored in objective, market-observable conditions rather than shaped by opportunistic managerial bias. To mitigate these inherent risks, auditors obtain highly reliable external corroborating evidence not only from independent actuaries but also from plan asset trustees, such as major commercial banks and insurance companies managing the pension assets, and statutory records from the National Pension Service. Because both management and auditors typically rely on these objective third-party reports to determine the amount of defined benefit obligations, the boundaries of acceptable estimates are highly constrained, thereby reducing the likelihood of disagreement. Under the audit risk model, such actuarial complexity directly amplifies assertion-level inherent risk, requiring auditors to extend substantive procedures and critically assess assumptions against observable market benchmarks (Simunic, 1980; Houston et al., 1999).

However, it is equally important to note that the valuation of defined benefit obligations relies heavily on highly standardized actuarial reports prepared by independent external actuaries. This strict dependence on verifiable third-party evidence can substantially constrain managerial discretion, potentially mitigating the scope for auditor–client friction and giving rise to a negative association. Given the competing theoretical tension between the inherent complexity of pension accounting and the standardizing effect of external actuarial verification, the directional impact of defined benefit obligations on auditor–client disagreement remains an empirical question. Accordingly, we state the following non-directional hypothesis:

H4. 

Ceteris paribus, the level of net defined benefit obligations is systematically associated with auditor–client disagreement.

Non-financial firms commonly employ derivative financial instruments to hedge operational exposures, particularly foreign exchange and interest rate risks. However, the accounting treatment prescribed by IFRS 9 (Financial Instruments) is technically demanding, requiring rigorous prospective and retrospective effectiveness testing and meticulous documentation to qualify for hedge accounting. From an agency theory perspective, managers have strong incentives to use derivative instruments as a substitute for accrual-based earnings management, thereby smoothing reported earnings and deferring the recognition of adverse fair value changes (Barton, 2001; Pincus & Rajgopal, 2002; Choi et al., 2015).

Given the considerable complexity and significant litigation exposure associated with undetected derivative misstatements, auditors must strictly enforce fair value measurement standards and hedge accounting eligibility criteria. The structural tension between management’s incentive to stabilize reported earnings and the auditor’s obligation to enforce complex accounting rules generates substantial potential for disagreement. Under the audit risk model, the technical complexity of hedge accounting documentation and Level 2/3 derivative valuation structurally elevates inherent risk at the assertion level, requiring auditors to lower the acceptable level of detection risk and deploy specialist-assisted substantive procedures (Simunic, 1980; Ranasinghe et al., 2023). Accordingly, we hypothesize:

H5. 

Ceteris paribus, the level of derivative financial instruments is positively associated with auditor–client disagreement.

Synthesizing these hypotheses reflects a nuanced perspective on account-level risk: Traditional high-risk accounts with established audit protocols are expected to be associated with lower auditor–client disagreement, while newly emerging high-risk accounts are expected to be associated with higher disagreement.

3. Research Design

3.1. Empirical Research Model

To test the hypotheses developed in Section 2.3, we estimate the following empirical model:

GAP = α0 + α1INDEP + α2SIZE + α3GRW + α4LEV + α5ROA + α6LOSS + α7LIQ
+ α8CON + α9INTER + α10OUTER + α11BIG4 + α12TENURE + α13OWN
+ α14FOR + α15DATE + α16MK + ∑YR + ∑IND + ε

Dependent Variable. The dependent variable, GAP, captures auditor–client disagreement and is operationalized using two complementary measures. The first, DUMGAP, is a binary indicator equal to one if the absolute scaled difference between pre-audit and post-audit net income exceeds a specified materiality threshold, and zero otherwise. The second, ABSGAP, is the continuous absolute value of this scaled difference, capturing the magnitude of disagreement without regard to its direction:

ABSGAP = (|Pre-audit net income − Post-audit net income|/|Post-audit net income|)

DUMGAP = 1 if (|Pre-audit net income − Post-audit net income|/|Post-audit net income|) > 1%, otherwise 0

Pre-audit net income is manually collected from mandatory timely disclosures—specifically, filings titled “Sales Revenue or Profit Structure Variation of 30% or More (15% for Large Corporations)”—submitted to the Korea Exchange’s Electronic Disclosure System (KIND). These disclosures report a firm’s internally finalized financial results for the fiscal year prior to completion of the external audit. Post-audit net income is extracted from the corresponding audited annual financial statements archived in TS-2000. For each firm-year observation, pre-audit disclosures are matched to audited financial statements using the firm’s unique identifier, fiscal year-end (December 31), and disclosure date. It is important to note that under the revised Act on External Audit in Korea, auditors are strictly prohibited from providing any advisory assistance during the preparation of financial statements. Because management must submit unpolished pre-audit numbers independently, and auditors apply strict professional skepticism, even minor adjustments frequently cross the disclosure threshold, leading to a relatively high baseline frequency of formally observable gaps.

The following observations are excluded from the matched sample: (i) firm–years for which pre-audit net income cannot be reliably matched to post-audit net income; (ii) disclosures pertaining solely to a subsidiary rather than the consolidated or separate parent entity; and (iii) preliminary disclosures subsequently revised prior to the audit completion date, which renders the original pre-audit figure ambiguous as a baseline for disagreement measurement.

We set the initial DUMGAP threshold at 1% of post-audit net income, consistent with prior research. We acknowledge that this threshold is conservative relative to the regulatory disclosure trigger under the KIND system—which requires disclosure only when the profit structure changes by 30% or more for KOSPI firms (15% for KOSDAQ firms). This conservative threshold is intentional: it enables the capture of economically meaningful audit adjustments that fall below the mandatory disclosure threshold, including adjustments that are material in firms operating in low-profit environments. Nonetheless, to address potential concerns that this threshold may classify immaterial rounding differences as substantive disagreements, we conduct robustness tests using alternative thresholds and specifications in Section 4.4. ABSGAP is used as a continuous complement to DUMGAP, measuring the economic magnitude of the pre-audit to post-audit earnings gap without regard to whether the adjustment is upward or downward. Section 4.4 further disaggregates the sample by the sign of GAP to examine whether the hypothesized associations differ between income-increasing and income-decreasing audit adjustments.

This study is subject to several limitations that warrant acknowledgment and suggest directions for future research. First, the measurement approach does not fully distinguish between mandatory and voluntary pre-audit earnings disclosures under the KIND system. In principle, the determinants and implications of auditor–client disagreement may differ systematically between these two disclosure types: mandatory disclosures are triggered by threshold changes in profit structure, whereas voluntary disclosures may reflect more discretionary managerial choices. Comprehensively classifying disclosures by type would require extensive additional manual verification beyond the scope of the present study, and we leave a detailed analysis of this distinction to future research. Second, the research design does not fully address potential endogeneity concerns. Firms with high auditor–client disagreement may simultaneously exhibit weak ICFR, complex group structures, or distinctive governance characteristics that jointly determine both account balances and the likelihood of audit adjustment. In particular, investments in subsidiaries and associates may partially proxy for the organizational complexity of conglomerate firms rather than capturing pure account-level inherent risk. Although we control for a comprehensive set of observable firm characteristics, unobserved heterogeneity may introduce residual bias into the estimates. Future research could employ research designs that more directly address endogeneity—for instance, by exploiting exogenous shocks to account-level risk, constructing matched samples, or employing instrumental variable approaches—to further strengthen causal identification in this domain.

Variables of Interest. The primary variables of interest, INDEP, represent the relative magnitude of five accounts with high inherent audit risk, each scaled by total assets to facilitate cross-sectional comparability:

INDEP1 = Accounts receivable/Total assets

INDEP2 = Inventory/Total assets

INDEP3 = (Investments in subsidiaries + Investments in associates)/Total assets

INDEP4 = Net defined benefit obligations (Defined benefit obligation − Plan assets)/Total assets

INDEP5 = (Derivative assets + Derivative liabilities)/Total assets

A statistically significant positive coefficient on any INDEP variable indicates that a larger relative balance in that account is associated with a higher likelihood or greater magnitude of auditor–client disagreement. Conversely, a statistically significant negative coefficient indicates that a larger balance is associated with lower disagreement, which may reflect the availability of reliable corroborating evidence or the maturation of audit responses to that account’s risk profile.

Control Variables. The model incorporates a comprehensive set of control variables identified in prior research as determinants of auditor–client disagreement and audit quality. All continuous variables are winsorized at the 1st and 99th percentiles to mitigate the influence of extreme observations. Year and industry fixed effects are included in all specifications, and standard errors are clustered at the firm level to account for serial correlation. Variable definitions are provided in

Table 1. Variable Definitions.

Variable		Definition
GAP		Auditor–client disagreement; operationalized as DUMGAP or ABSGAP
	DUMGAP	= 1 if (|Pre-audit net income − Post-audit net income|/|Post-audit net income|) > 1%, otherwise 0;
	ABSGAP	(|Pre-audit net income − Post-audit net income|/|Post-audit net income|)
INDEP		Accounts with high audit risk
	INDEP1	Accounts receivable/Total assets
	INDEP2	Inventory/Total assets
	INDEP3	(Investments in subsidiaries + Investments in associates)/Total assets
	INDEP4	Net defined benefit obligations (Defined benefit obligation − Plan assets)/Total assets
	INDEP5	(Derivative assets + Derivative liabilities)/Total assets
SIZE		Natural logarithm of total assets
GRW		Total asset growth rate = (Total assetst − Total assetst−1)/Total assetst−1
LEV		Debt Ratio = Total liabilities/Total assets
ROA		Return on assets = Net income/Total assets (ROA is defined using net income to maintain strict theoretical consistency with the dependent variable (GAP), which is calculated based on the difference between pre-audit and post-audit net income.)
LOSS		=1 if net loss reported, otherwise 0
LIQ		Current ratio = Current assets/Current liabilities
CON		=1 if consolidated financial statements prepared, otherwise 0 (Importantly, the indicator for consolidated financial statements (CON) serves as a robust proxy for organizational and structural complexity, controlling for the baseline differences between complex conglomerate groups with subsidiaries and standalone entities)
INTER		Internal monitoring intensity = (Outside director + Audit committee) compensation/Total assets
OUTER		External monitoring intensity = Audit fees/Total assets
BIG4		=1 if audited by a Big 4 audit firm, otherwise 0
TENURE		Natural logarithm of auditor tenure (years) (Following standard econometric practice, this study uses the natural logarithm of auditor tenure to account for the diminishing marginal impact of the auditor–client relationship on audit quality over time.)
OWN		Largest shareholder ownership ratio
FOR		Foreign shareholder ownership ratio
DATE		Natural logarithm of the number of days from fiscal year-end to pre-audit earnings disclosure date
MK		=1 if listed on KOSPI, otherwise 0 (KOSDAQ)

. While incorporating firm fixed effects could theoretically control for unobserved time-invariant heterogeneity, the sample period is relatively short (2021–2023), and the primary independent variables-account balances scaled by total assets-are highly sticky, exhibiting minimal within-firm variation over this brief window. Employing firm fixed effects would inappropriately absorb the critical cross-sectional variation necessary to test the hypotheses. Therefore, consistent with standard archival audit literature, the model relies on industry and year fixed effects to control for macroeconomic shocks and unobserved industry-level characteristics. To mitigate the influence of extreme observations and potential skewness, all continuous variables were winsorized at the 1st and 99th percentiles.

3.2. Data and Sample Selection

The initial sample comprises all non-financial firms listed on the Korea Exchange (KOSPI and KOSDAQ) over the period 2021–2023. The selection of this sample period is both theoretically and institutionally motivated. The sample period begins in 2021 to capture the post-stabilization phase of the comprehensive revision of the External Audit Act (implemented in 2018). By 2021, stringent regulatory measures such as standard audit hours and mandatory auditor designation were fully embedded, providing a stable institutional setting free from the confounding noise of the regulatory transition period. Following the comprehensive revision of the Act on External Audit of Stock Companies (the “New External Audit Act”), a suite of stringent regulatory measures—including the standard audit hour requirement and the mandatory auditor rotation and designation system—became fully operative and institutionally embedded by 2021. Prior literature documents that this regulatory overhaul fundamentally restructured the Korean audit environment by structurally reinforcing auditor independence, elevating professional conservatism, and promoting the rigorous application of risk-based audit frameworks (M. G. Lee et al., 2020; E. Lee & Park, 2016). By confining the sample to the post-stabilization period of 2021–2023, this study ensures a homogeneous and highly regulated institutional setting, thereby mitigating the confounding influence of ongoing regulatory transition on the observed association between account-level inherent risk and auditor–client disagreement. In addition, the final sample of 3852 firm-year observations provides sufficient statistical power to detect account-level effects and is broadly comparable to, or exceeds, the sample sizes employed in recent archival studies examining audit adjustments in the Korean capital market (e.g., Kim, 2024; Sohn et al., 2010). The following sequential exclusion criteria are applied to construct the final sample:

• The firm maintains a December 31 fiscal year-end, ensuring a uniform reporting cycle and eliminating confounding variation in disclosure timing across observations.
• The firm filed at least one mandatory timely disclosure—specifically, a “Sales Revenue or Profit Structure Variation” filing on the Korea Exchange’s Electronic Disclosure System (KIND)—during the sample period, enabling the manual collection of pre-audit net income. Disclosures filed solely by subsidiaries or business segments, rather than by the consolidated reporting entity, are excluded.
• Audited financial statement data required for the construction of all dependent, independent, and control variables are available from the TS-2000 database.

To construct the disagreement measure, we first identify all timely disclosures reporting preliminary financial results for each fiscal year in the sample period. For each disclosure, we manually extract the reported pre-audit net income figure and verify the identity of the reporting entity, the relevant fiscal year, and the disclosure date. We then match each verified disclosure to the corresponding firm-year observation in TS-2000 using the firm’s unique identifier and fiscal year-end date to obtain post-audit net income from the audited annual financial statements. Observations for which pre-audit net income cannot be reliably matched to post-audit net income, or for which key variable data—including high-inherent-risk account balances—are missing or internally inconsistent, are excluded from the final sample. Starting with an initial sample of 6878 firm-year observations listed on the Korea Exchange between 2021 and 2023, I applied several standard exclusion criteria to construct the final dataset. First, I excluded 482 observations from financial and insurance industries, as their distinct regulatory environments and financial statement structures limit comparability with non-financial firms. Second, I excluded 2544 observations that lacked the sufficient pre-audit or post-audit financial data required to compute the dependent and control variables. This screening process resulted in a final sample of 3852 firm-year observations for the primary empirical analysis.

To mitigate the influence of extreme observations, all continuous variables are winsorized at the 1st and 99th percentiles. The final sample consists of 3852 firm-year observations. Descriptive statistics for all variables are presented in

Table 2. Descriptive Statistics (N = 3852).

Variable	Mean	Std. Dev	P1	Median	P99
DUMGAP	0.6181	0.4859	0	1	1
ABSGAP	0.1494	0.4091	0	0.0122	2.5456
INDEP1	0.1430	0.0970	0.0022	0.1246	0.4467
INDEP2	0.0639	0.0892	0	0.0042	0.3635
INDEP3	0.0051	0.0300	0	0	0.1834
INDEP4	0.0088	0.0171	0	0	0.0827
INDEP5	0.0017	0.0079	0	0	0.0400
SIZE	26.1971	1.4684	23.7408	25.8934	31.1328
GRW	0.0985	0.3757	−0.5427	0.0404	1.8535
LEV	0.4326	0.2025	0.0567	0.4424	0.8636
ROA	0.0031	0.1062	−0.3989	0.0169	0.2596
LOSS	0.3448	0.4754	0	0	1
LIQ	0.7726	0.5652	0.0638	0.6657	2.9664
CON	0.8178	0.3861	0	1	1
INTER	0.0005	0.0005	0.0000	0.0003	0.0029
OUTER	0.0006	0.0006	0.0000	0.0004	0.0035
BIG4	0.4712	0.4992	0	0	1
TENURE	1.3282	0.8738	0.0000	1.3863	2.7408
OWN	0.3891	0.1700	0.0695	0.3809	0.7869
FOR	0.0680	0.1046	0.0000	0.0258	0.5224
DATE	3.7867	0.2935	2.9957	3.7842	4.2905
MK	0.3876	0.4873	0	0	1

, and Pearson correlation coefficients among the primary variables are reported in

Table 3. Correlations (N= 3852).

Variables		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)	(17)	(18)	(19)	(20)	(21)
(1)	DUMGAP
(2)	ABSGAP	0.29
(3)	INDEP1	−0.07	0.00
(4)	INDEP2	−0.05	−0.04	0.15
(5)	INDEP3	0.05	0.05	−0.03	0.04
(6)	INDEP4	−0.07	−0.01	0.11	0.34	0.05
(7)	INDEP5	0.05	0.01	0.00	0.10	0.01	0.06
(8)	SIZE	−0.11	−0.06	−0.10	−0.15	−0.09	−0.16	−0.06
(9)	GRW	0.05	0.02	0.04	0.00	−0.02	−0.02	0.06	0.01
(10)	LEV	0.10	0.12	0.15	0.02	−0.07	0.13	0.11	0.29	0.05
(11)	ROA	−0.11	−0.09	0.08	0.05	−0.04	−0.05	−0.10	0.25	0.17	−0.18
(12)	LOSS	0.11	0.15	−0.06	−0.04	0.02	0.04	0.08	−0.23	−0.15	0.16	−0.68
(13)	LIQ	0.07	0.10	−0.12	−0.10	0.03	0.05	0.03	0.18	−0.02	0.66	−0.16	0.15
(14)	CON	0.06	0.03	0.02	−0.23	−0.36	−0.27	−0.02	0.26	0.02	0.17	0.00	0.03	0.05
(15)	INTER	0.00	0.00	0.05	0.06	0.11	0.14	0.07	−0.51	−0.03	−0.16	−0.23	0.17	−0.09	−0.20
(16)	OUTER	0.11	0.09	0.02	0.03	0.06	0.17	0.09	−0.63	−0.03	−0.09	−0.40	0.30	−0.03	−0.14	0.55
(17)	BIG4	−0.03	0.00	−0.03	−0.04	−0.05	0.00	−0.06	0.42	−0.03	0.09	0.11	−0.11	0.04	0.08	−0.18	−0.14
(18)	TENURE	−0.11	−0.02	0.06	0.00	0.01	−0.04	−0.08	0.09	−0.04	−0.10	0.07	−0.08	−0.06	0.01	−0.06	−0.17	0.09
(19)	OWN	−0.11	−0.10	−0.02	0.08	0.00	0.03	−0.11	0.19	−0.04	−0.09	0.28	−0.25	−0.02	−0.09	−0.19	−0.29	0.21	0.04
(20)	FOR	−0.11	−0.08	−0.06	−0.08	−0.08	−0.08	−0.03	0.49	0.04	−0.05	0.22	−0.20	−0.08	0.14	−0.18	−0.24	0.27	0.11	−0.03
(21)	DATE	−0.05	0.02	−0.02	−0.05	−0.03	−0.05	0.00	−0.18	−0.01	0.12	−0.17	0.19	0.10	0.19	0.03	0.10	−0.15	−0.03	−0.10	−0.17
(22)	MK	−0.08	−0.02	−0.01	0.00	−0.05	0.03	−0.07	0.54	−0.04	0.16	0.12	−0.15	0.11	0.05	−0.26	−0.33	0.28	0.06	0.22	0.23	−0.22

Notes: The results in Table 3 are Pearson correlation coefficients. Bold indicates significance at the 1%, 5%, and 10% levels. Variable descriptions are as provided in Table 1.

.

4. Empirical Results

4.1. Descriptive Statistics

Table 2 reports descriptive statistics for all variables employed in the main analysis. All continuous variables are winsorized at the 1st and 99th percentiles to mitigate the influence of extreme observations.

Dependent Variables. The mean value of DUMGAP is 0.6181, indicating that 61.81% of firm-year observations in the sample exhibit a measurable discrepancy between pre-audit and post-audit net income exceeding the 1% materiality threshold. The mean value of ABSGAP is 0.1494, implying that, on average, post-audit net income deviates from pre-audit net income by approximately 14.94%. The prevalence and magnitude of these adjustments underscore the economic significance of auditor–client disagreement in the Korean listed firm context and validate the empirical relevance of the disclosure-based measurement approach adopted in this study.

Variables of Interest. The mean value of INDEP1 (accounts receivable scaled by total assets) is 0.1430, indicating that accounts receivable constitute approximately 14.30% of total assets on average—consistent with prior evidence that this account represents a material component of the balance sheet for non-financial firms. The mean value of INDEP2 (inventory scaled by total assets) is 0.0639, reflecting an average inventory-to-assets ratio of 6.39%. The mean values of INDEP3 (investments in subsidiaries and associates), INDEP4 (net defined benefit obligations), and INDEP5 (derivative financial instruments) are 0.0051, 0.0088, and 0.0017, respectively, indicating that these accounts constitute relatively modest proportions of total assets on average—approximately 0.51%, 0.88%, and 0.17%, respectively.

Notably, the median values of INDEP3, INDEP4, and INDEP5 are zero, indicating that a substantial proportion of firm-year observations carry no balance in these accounts. To ensure that the primary inferences are not confounded by the prevalence of zero observations, we re-estimate the main regression model on subsamples restricted to firm-years reporting non-zero balances for each respective account. The untabulated results indicate that the direction and statistical significance of the coefficients on INDEP3 through INDEP5 are qualitatively consistent with those reported in the main analysis, although statistical significance is attenuated in certain specifications, reflecting the reduction in sample size. These findings confirm that the main results are not an artifact of by a disproportionate concentration of zero-balance observations.

Control Variables. The mean value of SIZE is 26.1971, corresponding to average total assets of approximately KRW 238.3 billion. The mean values of GRW and LEV are 0.0985 and 0.4326, indicating average total asset growth of 9.85% and an average debt-to-assets ratio of 43.26%, respectively. The mean value of ROA is 0.0031, reflecting a modest average return on assets of 0.31% across the full sample. The mean value of LOSS is 0.3448, indicating that approximately 34.48% of firm-year observations reported a net loss during the sample period—a proportion that underscores the financial heterogeneity of the sample and motivates the inclusion of LOSS as a control variable. The mean (median) value of LIQ is 0.7726 (0.6657), and the mean value of CON is 0.8178, indicating that approximately 81.78% of observations are based on consolidated financial statements.

The mean values of INTER and OUTER—proxies for internal and external monitoring intensity, respectively—are 0.0005 and 0.0006, both scaled by total assets. The mean value of BIG4 is 0.4712, indicating that approximately 47.12% of firm-year observations are audited by a Big 4 audit firm. The mean values of OWN and FOR are 0.3891 and 0.0680, reflecting average largest-shareholder and foreign ownership ratios of 38.91% and 6.80%, respectively. The mean value of DATE is 3.7867 in logarithmic terms, corresponding to approximately 44 calendar days between the fiscal year-end and the mandatory pre-audit earnings disclosure date. Finally, the mean value of MK is 0.3876, indicating that approximately 38.76% of the sample consists of KOSPI-listed firms, with the remainder listed on KOSDAQ.

4.2. Correlations

Table 3 reports Pearson correlation coefficients among the primary variables. The correlation between the two dependent variables, DUMGAP and ABSGAP, is positive and statistically significant, confirming that the incidence and magnitude of auditor–client disagreement are positively related—that is, firm-years more likely to exhibit a discrepancy between pre-audit and post-audit net income also tend to exhibit larger absolute adjustments.

Correlations with Variables of Interest. DUMGAP exhibits statistically significant negative correlations with INDEP1, INDEP2, and INDEP4, indicating that larger relative balances in accounts receivable, inventory, and net defined benefit obligations are associated with a lower likelihood of auditor–client disagreement. These univariate patterns are directionally consistent with the main regression results and provide preliminary support for the interpretation that mature audit responses and the availability of reliable external corroborating evidence attenuate disagreement in these accounts. The correlations between ABSGAP and the variables of interest are directionally similar, though they differ in the level of statistical significance across specific accounts.

Correlations with Control Variables. DUMGAP exhibits a statistically significant negative correlation with SIZE, consistent with the view that larger firms are subject to greater external scrutiny and maintain more sophisticated internal reporting processes, which reduce the scope for material pre-audit to post-audit earnings adjustments. DUMGAP is positively and significantly correlated with GRW, suggesting that high-growth firms—which often face more complex accounting judgments and elevated inherent risk across multiple accounts—are more likely to experience contested audit adjustments. The positive and significant correlation between DUMGAP and LEV is consistent with prior research demonstrating that higher financial leverage intensifies debt covenant pressure and management’s incentives to resist downward audit adjustments (Yoon, 2001).

DUMGAP exhibits a statistically significant negative correlation with ROA, indicating that more profitable firms are less likely to experience auditor–client disagreement, consistent with the view that stronger operating performance reduces management’s incentive to engage in aggressive reporting. Conversely, DUMGAP is positively and significantly correlated with LOSS, suggesting that firms reporting a net loss face a higher likelihood of disagreement—a pattern consistent with prior evidence that audit adjustments increase in magnitude when firms report losses, as management has stronger incentives to resist income-reducing adjustments in loss environments. DUMGAP is also positively and significantly correlated with LIQ and CON, indicating that firms with higher current ratios and those preparing consolidated financial statements exhibit greater disagreement incidence, potentially reflecting the greater complexity of financial reporting in these settings.

DUMGAP exhibits statistically significant negative correlations with BIG4, TENURE, OWN, FOR, DATE, and MK. These patterns suggest that engagement of a Big 4 auditor, longer auditor tenure, higher ownership concentration among the largest shareholder, greater foreign investor participation, a later pre-audit disclosure date relative to fiscal year-end, and listing on the KOSPI market are each associated with a lower likelihood of auditor–client disagreement. The correlations between ABSGAP and the control variables are directionally consistent with those reported for DUMGAP, though statistical significance varies across individual variables.

We note that the pairwise correlations among the independent and control variables are generally of moderate magnitude, with no coefficient approaching levels that would raise concerns about multicollinearity in the regression specifications. Formal variance inflation factor (VIF) diagnostics, reported in conjunction with the regression results in

Table 4. Main Regression Results: High-Inherent-Risk Accounts and Auditor–Client Disagreement (n = 3852).

DUMGAP(ABSGAP)t = α0 + α1INDEP + α2SIZEt + α3GRWt + α4LEVt + α5ROAt + α6LOSSt + α7LIQt + α8CONt + α9INTERt + α10OUTERt + α11BIG4t + α12TENUREt + α13OWNt + α14FORt + α15DATEt + α16MKt + ∑YR + ∑IND + ε

Panel A. Logistic Regression (Dependent = DUMGAP)

Variables

Coef.

Chi

Coef.

Chi

Coef.

Chi

Coef.

Chi

Coef.

Chi

Coef.

Chi

Intercept

12.08

77.73 ***

11.38

69.80 ***

10.67

62.76 ***

11.36

70.13 ***

10.35

59.86 ***

13.82

95.70 ***

INDEP1

−2.80

45.76 ***

−2.79

43.85 ***

INDEP2

−1.93

19.21 ***

−1.20

6.61 ***

INDEP3

7.79

25.43 ***

7.78

24.79 ***

INDEP4

−14.15

39.11 ***

−12.06

26.01 ***

INDEP5

3.47

0.42

5.42

0.91

SIZE

−0.28

35.69 ***

−0.26

30.91 ***

−0.24

27.90 ***

−0.26

31.76 ***

−0.23

24.45 ***

−0.34

51.52 ***

GRW

0.34

9.87 ***

0.33

9.67 ***

0.33

9.92 ***

0.32

9.05 ***

0.33

9.44 ***

0.33

9.12 ***

LEV

1.58

30.68 ***

1.09

16.79 ***

0.95

13.11 ***

1.27

22.18 ***

0.84

10.33 ***

2.14

50.71 ***

ROA

−0.52

1.01

−0.72

1.95 *

−0.75

2.12 *

−0.67

1.71 *

−0.76

2.23 *

−0.35

0.45

LOSS

0.12

1.34

0.14

1.73 *

0.14

1.73 *

0.15

2.00 *

0.14

1.93 *

0.11

1.06

LIQ

−0.07

0.65

0.04

0.19

0.08

0.86

0.07

0.56

0.11

1.51

−0.17

3.31 **

CON

0.63

38.59 ***

0.50

24.34 ***

0.79

55.36 ***

0.44

18.02 ***

0.59

34.84 ***

0.65

34.08 ***

INTER

−410

25.48 ***

−424

27.44 ***

−444

29.54 ***

−406

25.05 ***

−421

27.19 ***

−430

26.72 ***

OUTER

206

4.64 **

202

4.47 **

215

5.02 **

263

7.39 ***

211

4.91 **

251

6.53 ***

BIG4

0.22

7.43 ***

0.24

8.37 ***

0.23

8.11 ***

0.24

8.43 ***

0.22

7.41 ***

0.26

9.82 ***

TENURE

−0.16

15.16 ***

−0.19

20.55 ***

−0.20

21.82 ***

−0.19

21.17 ***

−0.19

20.27 ***

−0.17

16.41 ***

OWN

−1.00

17.28 ***

−0.95

15.82 ***

−0.98

16.72 ***

−0.92

14.60 ***

−1.00

17.69 ***

−0.84

11.84 ***

FOR

−1.10

7.45 ***

−1.05

6.73 ***

−1.01

6.25 ***

−0.99

5.88 ***

−1.10

7.44 ***

−0.90

4.86 **

DATE

−1.16

73.23 ***

−1.12

69.11 ***

−1.15

71.33 ***

−1.13

70.16 ***

−1.10

66.76 ***

−1.24

80.64 ***

MK

−0.16

2.96 **

−0.18

3.76 **

−0.17

3.54 **

−0.14

2.16 *

−0.19

4.20 **

−0.09

0.83

∑YR

Included

Included

Included

Included

Included

Included

∑IND

Included

Included

Included

Included

Included

Included

LR

401.71

374.55

387.63

395.06

355.86

478.45

Panel B. OLS Regression (Dependent = ABSGAP)

Variables

Coef.

t

Coef.

t

Coef.

t

Coef.

t

Coef.

t

Coef.

t

Intercept

0.97

3.91 ***

1.05

4.26 ***

0.97

3.99 ***

1.01

4.12 ***

0.95

3.89 ***

1.13

4.51 ***

INDEP1

−0.05

−0.60

−0.04

−0.56

INDEP2

−0.21

−2.61 ***

−0.16

−1.85 **

INDEP3

0.92

3.96 ***

0.91

3.89 ***

INDEP4

−0.99

−2.38 ***

−0.71

−1.64 *

INDEP5

−1.10

−1.32 *

−0.94

−1.13

SIZE

−0.03

−3.35 ***

−0.03

−3.67 ***

−0.03

−3.55 ***

−0.03

−3.56 ***

−0.03

−3.31 ***

−0.04

−4.05 ***

GRW

0.03

1.52 *

0.03

1.50 *

0.03

1.55 *

0.03

1.47 *

0.03

1.59 *

0.03

1.56 *

LEV

0.16

3.08 ***

0.17

3.54 ***

0.16

3.30 ***

0.17

3.55 ***

0.15

3.19 ***

0.21

4.02 ***

ROA

0.29

3.24 ***

0.30

3.30 ***

0.29

3.25 ***

0.30

3.30 ***

0.28

3.15 ***

0.31

3.40 ***

LOSS

0.13

6.89 ***

0.13

6.87 ***

0.13

6.86 ***

0.13

6.92 ***

0.13

6.91 ***

0.13

6.82 ***

LIQ

0.04

2.25 **

0.03

2.03 **

0.04

2.33 ***

0.04

2.34 ***

0.04

2.48 ***

0.03

1.57 *

CON

0.04

2.41 ***

0.03

1.85 **

0.07

3.59 ***

0.03

1.74 **

0.04

2.35 ***

0.06

2.70 ***

INTER

−45.5

−3.06 ***

−46.2

−3.11 ***

−48.5

−3.27 ***

−44.7

−3.01 ***

−45.3

−3.05 ***

−47.2

−3.18 ***

OUTER

32.1

2.04 **

31.2

1.99 **

32.4

2.06 **

35.1

2.23 **

32.5

2.06 **

33.8

2.15 **

BIG4

0.03

2.08 **

0.03

2.20 **

0.03

2.19 **

0.03

2.14 **

0.03

2.04 **

0.03

2.30 **

TENURE

0.01

0.68

0.00

0.64

0.00

0.55

0.00

0.62

0.00

0.56

0.00

0.56

OWN

−0.16

−3.60 ***

−0.15

−3.45 ***

−0.15

−3.50 ***

−0.15

−3.45 ***

−0.16

−3.66 ***

−0.15

−3.31 ***

FOR

−0.13

−1.63 *

−0.12

−1.56 *

−0.11

−1.48 *

−0.12

−1.53 *

−0.12

−1.61 *

−0.10

−1.34 *

DATE

−0.05

−1.97 **

−0.05

−2.02 **

−0.05

−2.08 **

−0.05

−2.01 **

−0.05

−1.99 **

−0.05

−2.25 **

MK

0.03

1.84 **

0.03

1.90 **

0.03

1.95 **

0.03

2.03 **

0.03

1.78 **

0.04

2.16 **

∑YR

Included

Included

Included

Included

Included

Included

∑IND

Included

Included

Included

Included

Included

Included

Adj. R2

0.0468

0.0484

0.0507

0.0482

0.0472

0.0523

Notes: Refer to <Empirical research model> for the variable definitions. ***, **, * indicates significance at the 1%, 5%, and 10% level, respectively (two-tailed). To check for multicollinearity, Variance Inflation Factors (VIF) were calculated. The highest VIF in the fully specified model is 2.85, which is well below the conventional threshold of 10, indicating no multicollinearity concerns.

, confirm that multicollinearity does not materially affect the estimation.

4.3. Main Analysis

Table 4 reports the results of estimating the research model across the five high-inherent-risk accounts. Panel A presents logit regression results using DUMGAP as the dependent variable; Panel B presents OLS regression results using ABSGAP. All specifications include year and industry fixed effects, and standard errors are clustered at the firm level.

(1)

Panel A: Logit Results (DUMGAP)

The coefficient on INDEP1 (accounts receivable scaled by total assets) is negative and statistically significant, indicating that firms with larger accounts receivable balances relative to total assets exhibit a lower likelihood of auditor–client disagreement. Similarly, the coefficient on INDEP2 (inventory scaled by total assets) is negative and statistically significant. These findings are consistent with the interpretation that accounts receivable and inventory—having long been recognized as audit-sensitive accounts—have been subject to sustained, intensive audit scrutiny over multiple engagement cycles. While not explicitly tested in our empirical model, one plausible theoretical explanation for this finding is the recent adoption of advanced audit technologies—such as automated confirmation platforms and drone-assisted inventory counts. These advancements may have standardized the verification procedures for these accounts, thereby structurally reducing the likelihood of auditor-client disagreements.

The coefficient on INDEP3 (investments in subsidiaries and associates scaled by total assets) is positive and statistically significant, indicating that larger relative balances in this account are associated with a higher likelihood of auditor–client disagreement. This result is consistent with H3 and reflects the interpretive uncertainty surrounding the impairment assessment and fair value estimation of subsidiary investments under K-IFRS—an area that has attracted increasing regulatory attention in recent enforcement actions. Because consensus on the appropriate accounting treatment for this account remains unsettled, auditors and management are more likely to hold divergent views, resulting in a higher incidence of pre-audit to post-audit earnings adjustments. It is noteworthy that this positive association holds robustly even after controlling for the firm’s overarching structural complexity through the consolidated financial statements indicator (CON) and overall size (SIZE). This confirms that the disagreement is strictly associated with by the magnitude of the valuation and impairment risk embedded in the account, rather than merely the organizational complexity of operating as a conglomerate.

The coefficient on INDEP4 (net defined benefit obligations scaled by total assets) is negative and statistically significant, indicating that firms with larger net defined benefit obligations exhibit a lower likelihood of auditor–client disagreement. This result is inconsistent with the direction predicted in H4; we attribute this negative association to the availability of highly reliable external corroborating evidence—specifically, actuarial reports prepared by independent actuaries, confirmations from external financial institutions, and pension data from the National Tax Service. The credibility of this third-party evidence effectively constrains the scope for management discretion and provides auditors with a robust basis for independent judgment, thereby limiting the frequency of contested adjustments despite the inherent complexity of defined benefit accounting under IAS 19.

The coefficient on INDEP5 (derivative financial instruments scaled by total assets) is not statistically significant in the logit specification, indicating that the relative magnitude of derivative positions is not associated with a dichotomous shift in the likelihood of auditor–client disagreement. These results are robust to the inclusion of all five INDEP variables in a single specification, with the signs and significance of INDEP1 through INDEP4 remaining qualitatively unchanged.

(2)

Panel B: OLS Results (ABSGAP)

The coefficient on INDEP1 is negative but does not attain conventional levels of statistical significance in the ABSGAP specification, suggesting that while larger accounts receivable balances are associated with a lower incidence of disagreement, their association with the magnitude of adjustment is less precisely estimated. The coefficient on INDEP2 remains negative and statistically significant, confirming that inventory is associated with both a lower likelihood and a lower magnitude of auditor–client disagreement—a result that reinforces the maturation-of-audit-response interpretation advanced for Panel A.

The coefficient on INDEP3 is positive and statistically significant, consistent with Panel A and confirming that investments in subsidiaries are associated with both a higher incidence and a greater magnitude of auditor–client disagreement. The coefficient on INDEP4 is negative and statistically significant, consistent with Panel A. The coefficient on INDEP5 is negative and statistically significant in the OLS specification. One plausible interpretation is that, while derivative positions are economically modest for the majority of sample firms (mean INDEP5 = 0.17% of total assets; median = 0) and therefore rarely trigger a binary shift from non-disagreement to disagreement, the high contractual specificity and availability of market-based valuations for derivative instruments may provide auditors with sufficient independent evidence to resolve valuation disputes efficiently when such positions are present—thereby reducing the magnitude of pre-audit to post-audit earnings adjustments. We acknowledge, however, that given the limited economic scale of derivatives in this sample, this inference should be treated with appropriate caution.

When all five INDEP variables are included simultaneously, the coefficients on INDEP2 and INDEP4 remain negative and statistically significant, while the coefficient on INDEP3 remains positive and statistically significant, confirming that the primary results are not attributable to collinearity among the variables of interest.

Control Variables. Across both panels, SIZE exhibits a negative and statistically significant coefficient, consistent with the view that larger firms face greater external scrutiny and maintain more sophisticated internal reporting infrastructure, both of which reduce the likelihood and magnitude of material pre-audit to post-audit earnings adjustments. The coefficients on GRW and LEV are positive and statistically significant in Panel A, indicating that high-growth firms—which face elevated inherent risk and complex accounting judgments—and highly leveraged firms—which face stronger incentives to resist downward audit adjustments under debt covenant pressure—are more likely to experience auditor–client disagreement. The positive and significant coefficient on CON reflects the greater reporting complexity associated with the preparation of consolidated financial statements, which increases the scope for pre-audit adjustments.

The coefficient on INTER (internal monitoring intensity) is negative and statistically significant, indicating that stronger internal governance reduces the frequency and magnitude of auditor–client disagreement—consistent with the view that effective internal oversight improves the quality of pre-audit financial reporting and narrows the gap between management’s initial assertions and auditors’ independent assessments. In contrast, the coefficient on OUTER (external monitoring intensity, proxied by audit fees scaled by total assets) is positive and statistically significant, suggesting that auditors who charge higher fees—reflecting greater perceived risk or effort—are also more likely to identify and require corrections, resulting in larger pre-audit to post-audit adjustments. The positive and significant coefficient on BIG4 is consistent with this interpretation: higher-quality Big 4 auditors exercise greater professional skepticism and demand more rigorous adjustments to financial statements prior to issuing an unmodified opinion.

We note that the positive coefficient on BIG4 in the multivariate specification contrasts with the negative univariate correlation reported in Table 3. This sign reversal is attributable to the inclusion of correlated firm-level controls—particularly SIZE, OUTER, and FOR, which are strongly positively correlated with Big 4 engagement—and is consistent with prior evidence that Big 4 auditors, after controlling for client characteristics, are associated with higher-quality audits that generate more rigorous corrections (DeAngelo, 1981).

The coefficients on TENURE, OWN, FOR, DATE, and MK are negative and statistically significant, indicating that longer auditor–client relationships, higher ownership concentration among the largest shareholder, greater foreign investor participation, later pre-audit disclosure dates, and KOSPI listing are each associated with a lower likelihood and smaller magnitude of auditor–client disagreement. The Panel B control variable results are directionally consistent with Panel A, with minor differences in statistical significance across individual variables.

Taken together, the results in Table 4 indicate that auditors respond heterogeneously to account-level risk. Traditional high-risk accounts with well-established audit protocols—accounts receivable, inventory, and net defined benefit obligations—are associated with lower auditor–client disagreement, reflecting either the maturation of audit responses over time or the availability of reliable external corroborating evidence. Conversely, investments in subsidiaries and associates—an account for which interpretive uncertainty remains unresolved—are associated with higher auditor–client disagreement. These findings suggest that regulatory and standard-setting attention should be directed toward accounts that are newly characterized as high-risk, where the absence of settled audit practice creates the greatest potential for contested adjustments.

4.4. Robustness: Alternative Definitions of Auditor–Client Disagreement

To assess the sensitivity of the main findings to the operationalization of auditor–client disagreement, we re-estimate the research model using two alternative specifications, reported in

Table 5. Robustness Check: Alternative Measures of Auditor–Client Disagreement.

Variables	REALGAP		DUMGAP_5%
	Coef.	t-Value	Coef.	z-Statistic
Intercept	−0.20	−0.35	−0.37	−0.76
INDEP1	−0.33	−0.23	0.57	1.19
INDEP2	8.51	1.16	−0.21	−0.35
Controls	Included		Included
ΣYR	Included		Included
ΣIND	Included		Included
N	3852		3852
R-squared	0.01 (Adj)		0.11 (Pseudo)

.

(1)

Signed Disagreement (REALGAP)

In the main analysis, ABSGAP captures the magnitude of the pre-audit to post-audit earnings adjustment without regard to its direction. To examine whether the hypothesized associations are attributable to by a systematic directional bias—that is, whether high-risk accounts are disproportionately associated with income-increasing or income-decreasing audit adjustments—we re-estimate the model using REALGAP, defined as the signed difference between pre-audit and post-audit net income scaled by the absolute value of post-audit net income. The coefficients on INDEP1 (accounts receivable) and INDEP2 (inventory) are not statistically significant in this specification. This finding is theoretically informative: it indicates that the negative association between these accounts and auditor–client disagreement documented in the main analysis reflects a reduction in the overall dispersion of audit adjustments, rather than a systematic tendency toward either upward or downward corrections. In other words, the primary channel through which traditional high-risk accounts attenuate disagreement is the narrowing of the absolute gap between management’s pre-audit assertions and auditors’ independent assessments, not a directional bias in the nature of those adjustments. This result reinforces the interpretation that accumulated audit experience with accounts receivable and inventory produces convergent expectations between auditors and management, reducing the magnitude of disagreement irrespective of its sign.

(2)

Higher Materiality Threshold (DUMGAP_5%)

The baseline DUMGAP specification employs a conservative 1% threshold to capture economically meaningful audit adjustments, including those that are material in low-profit environments. To address the concern that this threshold may classify immaterial rounding differences as substantive disagreements, we re-estimate the logit model using DUMGAP_5%, defined as one if the absolute scaled discrepancy between pre-audit and post-audit net income exceeds 5% of post-audit net income, and zero otherwise. Under this more stringent criterion, the coefficients on the primary variables of interest do not attain conventional levels of statistical significance. We attribute this result primarily to the substantial reduction in statistical power: audit adjustments exceeding 5% of net income represent relatively rare events in the sample, and the corresponding reduction in the number of DUMGAP_5% = 1 observations materially constrains the precision of the logit estimates. Importantly, the directional pattern of coefficients across INDEP1 through INDEP5 remains qualitatively consistent with the main analysis, suggesting that the loss of significance reflects limited power rather than a reversal of the underlying associations.

Taken together, the robustness tests in Table 5 indicate that the primary inferences are not an artifact of the particular operationalization of auditor–client disagreement employed in the main analysis. The main findings are most clearly identified when disagreement is measured as the absolute magnitude of the pre-audit to post-audit earnings gap, consistent with the theoretical argument that account-level inherent risk is associated with the extent of audit negotiation rather than its directional outcome. Furthermore, an audit adjustment amounting to 1% of total post-audit net income is economically highly material. In low-profit margin environments, such a variance can mean the difference between meeting or missing analyst forecasts and debt covenant thresholds. Any adjustment that survives the rigorous negotiation process and results in a formal pre-to-post audit gap represents a substantive divergence in accounting judgment, fully validating the 1% threshold.

4.5. Robustness: Abnormal Account Levels

To further assess the robustness of the main findings, we examine whether period-to-period changes in high-audit-risk account balances, rather than their levels, are associated with auditor–client disagreement. The motivation for this test is straightforward: if sudden, abnormal fluctuations in a specific account draw heightened auditor attention—for instance, by triggering professional skepticism regarding the completeness or valuation of a newly elevated balance—then change variables may capture a distinct dimension of account-level inherent risk that is not fully reflected in the balance-sheet level measures employed in the main analysis. Accordingly, for each account of interest, we construct a first-difference variable defined as ΔINDEP_i,t = INDEP_i,t − INDEP_i,t−1, capturing the year-over-year change in the account balance scaled by total assets.

The results, reported in

Table 6. Robustness Check: Abnormal Account Levels Using Change Variables.

Variables	DUMGAP (Logit)		ABSGAP (OLS)		REALGAP (OLS)
	Coef.	z-Statistic	Coef.	t-Value	Coef.	t-Value
Intercept	11.66 ***	3.59	0.19 ***	21.49	0.76 *	1.65
CHG_AR	0.052	0.06	−0.203	−0.97	−0.320	−1.47
CHG_INV	0.213	0.19	0.162	0.48	0.155	0.61
Controls	Included		Included		Included
ΣYR	Included		Included		Included
ΣIND	Included		Included		Included
N	2250		2715		2620
R-squared	0.13 (Pseudo)		0.18 (Adj)		0.18 (Adj)

Notes: All specifications are based on the research model and include the same set of control variables (unreported for brevity). Robust t-statistics clustered at the firm level are reported in parentheses. ***, and * indicate significance at the 1%, and 10% levels, respectively. The sample size in this table differs from the main analysis because the change-variable specifications require lagged observations.

, indicate that the coefficients on the change in accounts receivable (CHG_AR) and the change in inventory (CHG_INV) are not statistically significant in the ABSGAP specification. This finding carries an important inferential implication: it suggests that transitory, period-specific fluctuations in high-risk account balances are not the primary driver of auditor–client disagreement. Rather, the cumulative stock of exposure to account-level inherent risk—as captured by the relative magnitude of the balance-sheet level measures in the main analysis—is the more economically meaningful determinant of the frequency and magnitude of pre-audit to post-audit earnings adjustments. This interpretation is consistent with the theoretical argument that auditor–client disagreement arises from the structural characteristics of an account’s risk profile, namely its inherent estimation uncertainty, valuation complexity, and susceptibility to management discretion, which are features that accumulate and persist over time rather than arising episodically from year-to-year fluctuations.

Collectively, the results of this additional test corroborate the validity of the level-based measurement approach adopted in the main analysis and confirm that the primary findings are not an artifact of by transitory shocks to account balances that may coincidentally co-occur with audit adjustments.

4.6. Robustness: Presence of Specific Accounts

A potential concern with the main analysis is that the level-based measures for investments in subsidiaries and associates (INDEP3), net defined benefit obligations (INDEP4), and derivative financial instruments (INDEP5) contain a substantial proportion of zero observations, as documented in Section 4.1. To the extent that the estimated coefficients for these accounts in the main regressions are influenced by the binary distinction between firms that carry such balances and those that do not—rather than by variation in the magnitude of the balance among firms that do—the primary inferences may not accurately reflect the underlying economic relationship of interest. To address this concern, we construct binary indicator variables for each of the three accounts: D_SUB, equal to one if the firm carries a non-zero balance in investments in subsidiaries and associates; D_DB, equal to one if the firm reports a non-zero net defined benefit liability; and D_DER, equal to one if the firm carries any non-zero derivative asset or liability position. Each indicator is set to zero otherwise. We then replace the corresponding level variables with these presence dummies in an otherwise identical specification of the research model.

The results, reported in

Table 7. Presence of High-Risk Accounts.

Variables	ABSGAP (Model 1)		ABSGAP (Model 2)		ABSGAP (Model 3)
	Coef.	t-Value	Coef.	t-Value	Coef.	t-Value
Intercept	0.18 ***	23.97	0.18 ***	23.97	0.18 ***	23.97
D_SUB	−0.001	−0.01
D_DB			0.014	0.13
D_DER					0.002	0.02
Controls	Included		Included		Included
ΣYR	Included		Included		Included
ΣIND	Included		Included		Included
N	4352		4352		4352
Adj. R2	0.113		0.113		0.113

*** indicate significance at the 1% levels.

, indicate that the coefficients on D_SUB, D_DB, and D_DER are not statistically significant in the ABSGAP specifications. This finding suggests that the mere presence of these accounts—independent of their relative magnitude—does not systematically predict a higher likelihood or greater magnitude of auditor–client disagreement within the full sample. The absence of significance for the presence dummies is theoretically consistent with the account-level risk framework underlying this study: it is the scale of exposure to estimation uncertainty, valuation complexity, and management discretion—not simply the existence of an account—that determines the extent to which an account becomes a locus of audit negotiation. Accordingly, firms with nominally non-zero but economically negligible positions in these accounts are unlikely to experience the same degree of auditor scrutiny as firms with material balances.

These results provide an important validation of the primary measurement approach. They confirm that the statistically significant associations documented in the main analysis for the level-based INDEP variables are primarily related to by cross-sectional variation in the magnitude of account balances rather than by the binary presence or absence of those accounts. Consequently, the inclusion of zero-balance firm-year observations in the main sample does not introduce systematic bias into the primary inferences.

5. Discussion of the Findings

The empirical results of this study offer novel insights into the micro-level dynamics of auditor–client negotiations, extending the application of agency theory and the audit risk model from the firm level to the individual account level. Prior research on auditor–client disagreement has predominantly focused on firm-level determinants, including corporate governance structures (Sohn et al., 2010), internal control quality (Kim, 2024), and financial distress (DeFond & Jiambalvo, 1993). Although these studies establish the broader organizational and environmental drivers of reporting conflicts, they tend to treat the financial statements as a monolithic object of audit inquiry, thereby obscuring the account-specific mechanisms through which disagreements arise.

The findings of this study reveal that auditor–client disagreement is fundamentally heterogeneous across individual accounts, shaped by the interplay between the degree of estimation uncertainty and the maturity of established audit verification protocols. Contrary to the conventional assumption that all accounts carrying high inherent risk uniformly amplify audit friction, the evidence documents a theoretically meaningful asymmetry. For traditional high-risk accounts—specifically accounts receivable and inventory—the significant negative association with disagreement suggests that decades of accumulated audit practice, regulatory scrutiny, and technological advancement have fostered convergent expectations between management and auditors regarding acceptable accounting treatments. This pattern is consistent with a maturation of the audit response, wherein the boundaries of permissible managerial discretion are well-defined and consistently enforced, thereby preemptively constraining the scope for agency conflict before the formal audit process commences.

By contrast, the significant positive association between investments in subsidiaries and associates and auditor–client disagreement highlights a critical and growing vulnerability in contemporary financial reporting. From an agency theory perspective, managers are incentivized to exploit the unverifiable, forward-looking Level 3 inputs inherent in impairment testing—such as long-term growth projections and discount rate assumptions—to defer the recognition of economic losses (Ramanna & Watts, 2012). Because standard audit procedures face substantial objective limitations in independently verifying these highly subjective estimates, this account class emerges as a primary locus of auditor–client conflict, reflecting unresolved interpretive uncertainty within the prevailing K-IFRS impairment assessment regime (Griffith et al., 2015).

The significant negative association documented for net defined benefit obligations yields an additional theoretical insight, underscoring the role of independent third-party corroborating evidence in mitigating agency costs. When auditors are able to anchor their professional judgments on highly reliable external sources—including actuarial reports from independent actuaries and contribution records from tax and social security authorities—the scope for opportunistic managerial manipulation of actuarial assumptions is effectively constrained, neutralizing a substantial portion of the account’s inherent complexity.

Taken together, these findings advance a refined theoretical perspective on auditor–client disagreement: rather than representing a static by-product of overall firm-level risk, disagreement is more precisely characterized as a dynamic negotiation outcome jointly determined by an account’s susceptibility to managerial opportunism under agency theory and the availability of objective, externally verifiable audit evidence under the audit risk model. This account-level framework offers a more granular and empirically tractable lens for understanding the determinants of financial reporting adjustment outcomes than prior firm-level analyses have afforded.

6. Conclusions

This study examines the association between account-level audit risk and auditor–client disagreement among non-financial firms listed on the KOSPI and KOSDAQ markets from 2021 to 2023. Grounded in the audit risk model and agency theory, we analyze whether the relative magnitude of five accounts characterized by high inherent risk—accounts receivable, inventory, investments in subsidiaries and associates, net defined benefit obligations, and derivative financial instruments—is systematically associated with the likelihood and magnitude of auditor–client disagreement. Disagreement is measured as the scaled difference between pre-audit net income, manually collected from mandatory timely disclosures filed with the Korea Exchange’s Electronic Disclosure System (KIND), and post-audit net income reported in the audited financial statements.

The empirical analysis yields three principal findings. First, accounts receivable and inventory—accounts long recognized as carrying high inherent audit risk—exhibit a statistically significant negative association with auditor–client disagreement. We interpret this as a potential technology-driven effect: the adoption of advanced audit technologies and standardized verification procedures for these accounts may have constrained managerial discretion attenuating the frequency and magnitude of contested pre-audit to post-audit earnings adjustments. Second, investments in subsidiaries and associates—an account that has attracted increasing regulatory attention as an emerging source of material misstatement risk—exhibit a statistically significant positive association with auditor–client disagreement, consistent with unresolved interpretive uncertainty in impairment assessment and fair value estimation under K-IFRS. Third, net defined benefit obligations exhibit a statistically significant negative association with disagreement, which we attribute to auditors’ ability to obtain highly reliable external corroborating evidence from independent actuaries, external financial institutions, and tax authority records—evidence that effectively anchors auditor judgment and constrains the scope for management discretion despite the inherent complexity of defined benefit accounting under IAS 19.

These findings collectively demonstrate that auditors respond heterogeneously to account-level inherent risk, and that this differential response is directly observable in the pre-audit to post-audit earnings gap. Specifically, the results reveal a theoretically meaningful asymmetry: traditional high-risk accounts with mature, well-established audit protocols are associated with lower auditor–client disagreement, whereas a newly prominent, complex account—investments in subsidiaries and associates—is associated with higher disagreement. This asymmetry suggests that the locus of audit negotiation shifts over time as audit practice evolves in response to emerging risk areas, and that regulatory and standard-setting attention should be directed toward accounts for which interpretive consensus has yet to be established.

The findings of this study carry substantive theoretical and practical implications. From a theoretical standpoint, this study advances the extant literature by integrating agency theory and the audit risk model at the individual account level. Whereas prior archival research has predominantly examined auditor–client friction through the lens of firm-level characteristics, this study provides direct empirical evidence that the locus of disagreement is dynamically determined by the specific inherent risk profile, valuation complexity, and external verifiability of individual balance sheet accounts. This account-level analytical framework bridges a notable gap in the literature by offering a more precise and empirically tractable theoretical lens for understanding how managerial opportunism and auditor conservatism interact during the pre-issuance financial reporting process—insights that aggregate, firm-level analyses are inherently unable to surface.

From a practical standpoint, the results yield actionable implications for audit practitioners, regulators, and standard-setters. For audit practitioners, the finding that newly prominent and structurally complex accounts—most notably investments in subsidiaries and associates—are associated with significantly elevated disagreement underscores the imperative of dynamically reallocating audit resources in response to evolving risk landscapes. Engagement partners should proactively deploy specialized personnel, including independent valuation specialists, to these high-friction areas during the engagement planning phase, thereby resolving interpretive uncertainties before they escalate into material audit adjustments. For regulators and standard-setters, the documented asymmetry in disagreement incidence across account categories suggests that regulatory oversight and enforcement efforts should not apply uniform scrutiny across all designated high-risk accounts. Rather, regulatory attention should be systematically directed toward accounts characterized by persistent interpretive uncertainty, limited external verifiability, and elevated susceptibility to managerial reporting bias under the prevailing K-IFRS measurement framework.

Finally, our study is subject to certain endogeneity concerns. While the empirical model controls for various firm characteristics and incorporates industry and year fixed effects, account balances are not strictly exogenous. Because the short panel period precludes the effective use of firm fixed effects—which would absorb the sticky within-firm variation of account balances—potential omitted variable bias remains. Consequently, the empirical findings should be interpreted cautiously as robust associations rather than definitive causal relationships. While the model controls for various firm characteristics, account balances are not strictly exogenous. Due to the short panel period (2021–2023), firm fixed effects could not be effectively employed without absorbing critical cross-sectional variation. Therefore, the findings should be interpreted cautiously as robust associations rather than definitive causal relationships.