0 cumulative citations
View corpus contextAudits still miss material fraud despite tighter rules; integrating behavioural reforms, stronger incentives and AI analytics is necessary to narrow the gap, though tech adoption risks concentrate audit quality and raise governance challenges.
Bridging the expectation performance gap in fraud detection: A systematic review of external auditors' responsibilities, audit failures, and regulatory responses
openalex
review_meta
medium evidence
7/10 relevance
Summary only
summary available; pdf_status=paywall
DOI
Source
PDF
A systematic review finds a persistent expectation–performance gap in external audit fraud detection and argues that closing it requires combining behavioural interventions, stronger institutional safeguards, and advanced analytics/AI rather than relying on compliance checklists alone.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
0 cumulative citations
View corpus contextFinancial statement fraud remains a persistent threat to the integrity of global capital markets despite extensive auditing standards and regulatory reforms. This study systematically reviews literature on external auditors' responsibilities in fraud detection and evaluates why significant audit failures continue to occur. Following the Preferred Reporting Items for Systematic Reviews and Meta Analyses framework, 1,232 studies were initially identified from Google Scholar and Crossref, of which 32 studies published between 2002 and 2025 met the inclusion criteria. The findings reveal a persistent expectation and performance gap between auditing standards, particularly ISA 240, and actual audit practice. Although auditors are required to obtain reasonable assurance that financial statements are free from material misstatement due to fraud, high profile corporate failures such as Enron, Wirecard, Carillion and Steinhoff demonstrate recurring deficiencies in fraud detection. The synthesis identifies four dominant factors influencing audit effectiveness: regulatory environment, auditor independence, professional competence and scepticism, and the complexity of fraud schemes. The review further shows that audit failures are driven not only by technical limitations but also by behavioural biases, institutional pressures and increasing fraud sophistication. Regulatory reforms such as the Sarbanes Oxley Act and CLERP 9 have improved compliance but have not eliminated audit deficiencies. The study contributes a multidimensional conceptual framework integrating behavioural, institutional and technological perspectives to explain persistent audit failures. It concludes that improving fraud detection requires a shift from compliance based auditing toward a more integrated approach that combines behavioural insight, enhanced professional training and advanced audit technologies such as artificial intelligence and data analytics.
Summary
Main Finding
Despite strengthened standards and reforms, a persistent expectation–performance gap remains: external audits often fail to detect material fraud. Audit effectiveness is shaped by regulatory settings, auditor incentives and competencies, behavioural biases, and increasing fraud complexity. The review argues that closing the gap requires moving beyond compliance checklists toward integrated approaches that combine behavioural insights, enhanced training, and advanced audit technologies (notably AI and data analytics).
Key Points
- Scope and evidence: Systematic review of 1,232 initially identified studies (Google Scholar, Crossref), with 32 studies (2002–2025) meeting inclusion criteria following PRISMA procedures.
- Expectation–performance gap: ISA 240 and similar standards require reasonable assurance against fraud, but practice frequently falls short—as illustrated by Enron, Wirecard, Carillion, Steinhoff.
- Four dominant drivers of audit failure:
- Regulatory environment (enforcement intensity, inspection regimes, legal liability).
- Auditor independence (economic ties, tenure, non‑audit fees).
- Professional competence and scepticism (training, experience, cognitive biases).
- Complexity and sophistication of fraud schemes (use of opaque instruments, related‑party transactions, technology).
- Non‑technical causes: Behavioural biases (overreliance, confirmation bias), institutional pressures (client retention, billable hours), and incentives explain many failures beyond pure technical limits.
- Regulatory reforms: Measures such as Sarbanes–Oxley and CLERP 9 increased compliance and controls but have not eliminated failures—often because reforms do not change behavioural incentives or keep pace with fraud innovation.
- Conceptual contribution: A multidimensional framework integrating behavioural, institutional and technological perspectives to explain persistent audit failures.
- Recommended remedy: Transition from compliance‑based auditing to integrated models that combine (i) behavioural interventions (training, incentives), (ii) stronger institutional safeguards, and (iii) advanced technologies (AI, analytics) to detect complex anomalies.
Data & Methods
- Review protocol: PRISMA framework used to identify, screen and select studies.
- Search sources: Google Scholar and Crossref for relevant literature (initial N = 1,232).
- Inclusion criteria: Empirical and conceptual papers addressing external auditors’ fraud detection responsibilities and audit failures; publication window 2002–2025.
- Final sample: 32 studies synthesized qualitatively.
- Synthesis approach: Thematic coding to identify recurring factors, case analyses of high‑profile failures, and integration into a multidimensional conceptual framework. (No meta‑analytic statistical pooling reported.)
Implications for AI Economics
- Role for AI and analytics
- Potential: AI can materially improve fraud detection by scanning large transaction datasets, uncovering non‑linear patterns, and flagging anomalies that human auditors miss.
- Complementarity: AI tools are likely complementary to auditor expertise—improving productivity, coverage, and the effective exercise of professional scepticism.
- Economic impacts on the auditing market
- Labor demand: AI adoption may shift skill demand toward data science, model governance, and interpretability skills; routine audit tasks may be automated.
- Competition and market structure: Large firms with access to superior data and AI capabilities may widen quality gaps; smaller firms face adoption barriers and potential market exit or consolidation.
- Pricing and liability: AI could change audit cost structures and risk allocation—affecting fee setting, insurance, and legal exposure.
- Incentives and strategic responses
- Endogenous fraud sophistication: As AI detection improves, fraudsters may adapt (adversarial strategies), raising an arms‑race dynamic; models must account for strategic complementarities between detection technology and fraud innovation.
- Moral hazard and overreliance: Excess trust in AI outputs risks automation bias; auditors may reduce scepticism if governance and explainability are weak.
- Policy and regulatory considerations
- Standards for AI use: Need for regulation on model validation, transparency, audit trails, and independence of AI tool providers.
- Data access and privacy: Effective AI auditing requires granular data; regulatory frameworks must balance privacy, confidentiality, and public interest.
- Incentive design: Regulation should align auditor incentives (inspection, penalties, disclosures) to encourage effective use of AI and maintain professional scepticism.
- Research opportunities for AI economists
- Causal evaluation: Difference‑in‑differences or staggered rollouts to estimate the effect of AI audit tools on fraud detection, restatements, enforcement actions.
- Field experiments: Randomized trials assigning AI‑assisted audit modules to teams to measure changes in detection rates and behavioural responses.
- Structural models: Dynamic models of auditor–firm interactions capturing investment in AI, fraud effort, and enforcement regimes to study equilibrium outcomes.
- Market analysis: Study adoption heterogeneity, complementarities between AI and human skills, impacts on wages and firm entry/exit.
- Adversarial dynamics: Game‑theoretic or agent‑based models of adaptive fraudsters facing AI detectors.
- Data sources: PCAOB inspection reports, enforcement records, financial restatements, audit firm disclosures, transaction‑level datasets where available, and proprietary AI tool usage logs.
- Practical cautions
- Model risk and explainability: Black‑box models without interpretability can exacerbate legal and trust issues; economics research should weigh accuracy gains against governance costs.
- Distributional effects: AI adoption may concentrate auditing quality and increase inequality across firms and auditors; policy should consider transition support and standardization.
Suggested next steps for researchers: design empirical studies exploiting natural experiments in AI tool adoption at audit firms, collect panel data on detection outcomes, and collaborate with audit practitioners to access realistic transaction‑level datasets and to evaluate behavioural responses to AI support.
Assessment
Paper Typereview_meta
Evidence Strengthmedium — Systematic PRISMA-style screening yielded a focused qualitative synthesis (32 studies) spanning case analyses and empirical work; findings are consistent across sources but rely largely on observational, case-based, and conceptual studies without pooled quantitative estimates or causal identification.
Methods Rigormedium — Use of a PRISMA protocol and thematic coding indicates systematic effort, but searches were limited to Google Scholar and Crossref (no major databases listed), there is no reported risk-of-bias appraisal or meta-analytic pooling, and included studies are heterogeneous in design and quality.
SampleSystematic review: initial N=1,232 records identified via Google Scholar and Crossref; after screening and applying inclusion criteria (2002–2025) 32 studies were included and synthesized qualitatively using thematic coding plus illustrative case analyses of high-profile audit failures.
Themesgovernance human_ai_collab
GeneralizabilityFindings synthesize varied study designs and jurisdictions; applicability may differ by country and regulatory regime., Reliance on high-profile corporate scandals may overstate prevalence or mechanisms relative to routine audits., Search limited to Google Scholar and Crossref may have missed indexed literature in discipline-specific databases or non-English studies., Qualitative synthesis without meta-analysis limits ability to generalize effect sizes or causal magnitudes., Discussion of AI implications is largely conceptual and extrapolative pending causal, transaction-level evidence.
Claims (13)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| The review identified 1,232 studies initially and synthesized 32 studies published between 2002 and 2025 after applying PRISMA-based inclusion and screening procedures. Other | null_result | Scope and composition of the reviewed literature |
Reading fidelity
high
Study strength
high
|
n=1232
|
| External audits frequently fail to detect material fraud, leaving a persistent expectation–performance gap despite strengthened auditing standards and reforms. Error Rate | negative | External auditors' detection of material fraud |
Reading fidelity
high
Study strength
medium
|
n=32
|
| Audit effectiveness is shaped by regulatory settings, auditor incentives and competencies, behavioural biases, and the increasing complexity of fraud schemes. Error Rate | mixed | Audit effectiveness and fraud-detection performance |
Reading fidelity
high
Study strength
medium
|
n=32
|
| Behavioural biases and institutional pressures contribute to audit failures beyond purely technical limitations. Error Rate | negative | Auditor professional scepticism and fraud-detection effectiveness |
Reading fidelity
high
Study strength
medium
|
n=32
|
| Sarbanes–Oxley, CLERP 9, and similar reforms increased compliance and controls but did not eliminate audit failures. Regulatory Compliance | mixed | Audit-failure incidence and compliance/control strength |
Reading fidelity
high
Study strength
medium
|
n=32
|
| The review proposes a multidimensional framework integrating behavioural, institutional, and technological perspectives to explain persistent audit failures. Governance And Regulation | mixed | Conceptual explanation of audit failures |
Reading fidelity
high
Study strength
low
|
n=32
|
| The review recommends moving from compliance-based auditing toward integrated models combining behavioural interventions, institutional safeguards, and AI or data-analytics tools. Error Rate | positive | Fraud-detection effectiveness and audit quality |
Reading fidelity
high
Study strength
speculative
|
n=32
|
| AI and data analytics have the potential to improve fraud detection by scanning large transaction datasets, identifying non-linear patterns, and flagging anomalies that human auditors may miss. Error Rate | positive | Fraud-detection accuracy and anomaly identification |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| AI audit tools are likely to complement auditor expertise by improving productivity, coverage, and the effective exercise of professional scepticism. Organizational Efficiency | positive | Audit productivity, audit coverage, and professional scepticism |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| AI adoption in auditing may increase demand for data-science, model-governance, and interpretability skills while automating routine audit tasks. Task Allocation | mixed | Skill demand and automation of routine audit tasks |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| Unequal access to advanced AI capabilities may allow large audit firms to widen quality gaps, while smaller firms face adoption barriers and possible exit or consolidation. Market Structure | negative | Audit-market concentration and quality differences across firms |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| Improved AI-based fraud detection may induce fraudsters to adopt adaptive or adversarial strategies, creating an arms-race dynamic between detection technology and fraud innovation. Error Rate | mixed | Fraud sophistication and effectiveness of fraud detection |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| Overreliance on AI outputs can create automation bias and reduce auditors' professional scepticism when governance and explainability are weak. Decision Quality | negative | Auditor professional scepticism and judgment quality |
Reading fidelity
high
Study strength
speculative
|
not reported
|