Explaining the Key Determinants of Financial Fraud in Iranian Companies Through the Integration of Qualitative Data and the XGBoost Model

Abstract

This study aimed to identify, contextualize, validate, and rank the determinants of financial fraud risk in Iranian companies and to develop a context-sensitive model for fraud-risk assessment and prediction. A mixed-methods qualitative–quantitative approach with a descriptive survey design was employed. In the qualitative phase, 45 studies published during the preceding five years were examined through meta-synthesis, and semi-structured interviews were conducted with 12 experts possessing at least 10 years of professional experience in auditing, financial management, corporate governance, or capital-market regulation. Interview data were analyzed through thematic analysis using open, axial, and selective coding. Inter-coder reliability was 0.86. Integration of the qualitative findings produced an initial set of 17 determinants. In the quantitative phase, Pearson correlation coefficients and analysis of variance were applied in SPSS to assess the relationships between these determinants and financial fraud risk. Following statistical screening, 15 determinants were entered into the XGBoost algorithm and ranked using Gain Importance. Poor audit quality and weak board oversight demonstrated the strongest association with fraud risk (r=0.465, p<0.001; F=22.30, p<0.001). Significant relationships were also found for financial pressure and high leverage (r=0.421, p=0.001; F=18.92, p<0.001), fraud triangle/pentagon elements (r=0.403, p=0.001; F=16.54, p=0.001), and concentrated family ownership (r=0.398, p=0.002; F=16.01, p=0.001). Internal-auditor independence and operational complexity were excluded because their effects were statistically nonsignificant. The XGBoost ranking identified poor audit quality and weak board oversight (18.7%), financial pressure and high leverage (15.2%), and concentrated family ownership (12.4%) as the three most influential predictors, jointly accounting for nearly half of the model’s explanatory importance. Effective fraud-risk assessment in Iran requires the integration of universal indicators with context-specific economic, ownership, and regulatory conditions. Fraud-prevention policies should prioritize audit quality and independence, effective board oversight, ownership transparency, and systematic management of financial pressures.