Organisational scale and fraud volatility: An exploratory agent-based simulation of occupational fraud dynamics

Fraud is commonly studied as an individual-level phenomenon. Less attention has been given to how fraud behaves at the level of organisational systems. Empirical research on the relationship between organisational size and fraud has produced mixed findings, partly because detected-incident datasets cannot reveal the underlying generative mechanisms of fraud. This study addresses that gap by examining whether fraud scales systematically with organisational size, and whether social influence and network connectivity moderate that relationship. This study is an exploratory, conceptual agent-based simulation not empirically calibrated to real-world data. An agent-based model was developed in which employees interact within simulated organisations under varying structural and behavioural conditions. Simulations were run across nine organisational sizes (25-6,400 agents), ten levels of susceptibility to social influence (q_avg = 0 to 0.9), and five levels of network connectivity scaling (αk = 0-1), yielding 9,000 simulations in total. Log-log regression was used to estimate scaling exponents for three outcomes: mean fraud, peak fraud, and fraud volatility. Both average and peak fraud levels scale approximately linearly with organisational size. However, fraud volatility scales superlinearly (β = 1.393), meaning that a doubling of organisational size is associated with approximately 2.6 times greater variability in the level of fraud. Susceptibly to social influence and network connectivity significantly moderate the volatility of fraud activity but do not significantly affect mean fraud levels. Theoretically, these findings establish organisational scale as a structural mechanism shaping fraud dynamics and introduce fraud volatility as a distinct dimension of risk that scales disproportionately with size - one that average incidence measures fail to capture. Practically, proportional fraud control resourcing is adequate for managing average fraud levels, but managing volatility requires a different approach. Because fraud volatility is largely invisible in real organisations, the findings support reorienting fraud risk identification away from individual-centred frameworks toward affordance mapping - the systematic analysis of structural conditions that make fraud possible - as a more complete basis for directing counter-fraud effort than case-led detection alone.