Fraud dynamics and internal controls in organizations with localized interactions and norm formation.

摘要

This dissertation develops an agent-based model to consider how attitudes in an organization can evolve over time to support fraud and how internal controls can mitigate the resulting increase in fraud risk. It extends the work of Davis and Pesch (2011) by examining an alternative organizational model that emphasizes local interaction where agents interact primarily with those near them (e.g., in the same office or division) and social influence through the combined actions of one's peers (rather than one-to-one recruitment). The model consists of a group of heterogeneous agents and a set of simple social interaction rules that together define the organization. In accordance with Cressey's (1953) fraud triangle hypothesis, any agent encountering the union of motive, opportunity, and a pro-fraud attitude will commit fraud. Agents are allowed to repeatedly interact with neighboring agents, resulting in changes in agent attitudes toward fraud (via the cultural transmission of social norms).;The benchmark analysis revealed three general trends. First, regardless of agents' susceptibilities to social influence, the benchmark model tended toward a particular level of fraud (similar to a Lyapunov stable equilibrium). Second, as the agents' susceptibilities to influence became more homogeneous, the state distribution of aggregate fraud levels over time became more concentrated. Third, when average susceptibility to social influence was moderate to high, fraud risk was localized to subsets (i.e., departments or divisions) of the organization.;Next, variations of the model were developed to investigate how various internal control mechanisms—general controls, tone at the top, ethical training, and enforcement—impact the dynamics of fraud. Analysis of the internal control models suggests that the impact of mechanisms to prevent or eliminate fraud is contingent on the specific nature of social influence and the level of variation in individual agents' susceptibilities to social influence. The model analysis and results have implications for practitioners responsible for designing and testing internal control systems.