THE INFLUENCE OF STATISTICAL MODELS FOR FRAUD DETECTION IN PROCUREMENT AND INTERNATIONAL TRADE SYSTEMS

Abstract

Procurement and international trade systems remain particularly vulnerable to complex fraud mechanisms that undermine transparency, equity, and fiscal integrity across both public and private sectors. These fraudulent behaviors—ranging from misinvoicing and threshold bunching to split purchasing and restricted competition—continue to distort market signals, inflate transaction costs, and erode trust in institutional governance. Addressing these vulnerabilities requires not only the deployment of advanced analytic tools but also an understanding of how interpretable statistical modeling frameworks can enhance the credibility, precision, and utility of fraud detection outcomes in practice. This study therefore seeks to explore the empirical relationships among internal control mechanisms, vendor concentration patterns, compliance cultures, and the intensity of transactional anomalies within enterprise procurement operations. Employing a quantitative, cross-sectional, multi-case research design, this investigation draws upon survey and administrative data from five enterprise environments that operate either enterprise resource planning (ERP) or e-procurement systems. A total of 268 respondent-level observations were analyzed, providing a robust dataset for comparative evaluation. To anchor the analytical framework, a structured literature review of 37 peer-reviewed studies was conducted, synthesizing theoretical constructs from fraud risk modeling, internal audit research, and information system governance. The resulting conceptual model integrates both organizational and transactional dimensions, encompassing variables such as internal control strength, vendor risk exposure, concentration ratios, transaction anomaly intensity, compliance culture maturity, audit cadence, and a composite fraud-risk outcome index. The analytical sequence proceeded in a methodologically rigorous manner: preliminary reliability and validity diagnostics were performed to assess construct consistency; descriptive analyses established baseline patterns; followed by zero-order and partial correlation matrices to determine intervariable associations. Subsequently, fixed-effects regression modeling was implemented to isolate within-organization effects, employing ordinary least squares (OLS) for continuous outcomes, logistic regression for binary fraud event classifications, and negative binomial models for count-based incident outcomes. Moderation terms were prespecified to evaluate interaction effects between structural controls and transactional anomalies. The empirical findings reveal that transaction anomaly intensity constitutes the most powerful positive correlate of the composite fraud-risk score, confirming that irregular purchasing and invoicing behaviors remain key indicators of systemic exposure. Conversely, internal control robustness and compliance culture strength demonstrate significant negative relationships with fraud risk, illustrating their preventive potential. The results further indicate that vendor concentration, a proxy for limited competitive pressure—exerts a positive and statistically significant effect on risk, highlighting the importance of diversification strategies.